diff --git a/CHANGELOG b/CHANGELOG
index b8e5a8b539..24ac5b5da1 100644
--- a/CHANGELOG
+++ b/CHANGELOG
@@ -26,6 +26,8 @@ devel
   be used that has access to the `_system` database, in order to create
   the databases on restore. 
 
+* added index hints feature to AQL
+
 * added "name" property for indices
 
   If a name is not specified on index creation, one will be auto-generated.
diff --git a/Documentation/Books/AQL/Operations/For.md b/Documentation/Books/AQL/Operations/For.md
index a32ac8bd82..7f739225f5 100644
--- a/Documentation/Books/AQL/Operations/For.md
+++ b/Documentation/Books/AQL/Operations/For.md
@@ -33,7 +33,7 @@ required that *expression* returns an array in all cases. The empty array is
 allowed, too. The current array element is made available for further processing 
 in the variable specified by *variableName*.
 
-```
+```js
 FOR u IN users
   RETURN u
 ```
@@ -52,7 +52,7 @@ placed in is closed.
 
 Another example that uses a statically declared array of values to iterate over:
 
-```
+```js
 FOR year IN [ 2011, 2012, 2013 ]
   RETURN { "year" : year, "isLeapYear" : year % 4 == 0 && (year % 100 != 0 || year % 400 == 0) }
 ```
@@ -61,7 +61,7 @@ Nesting of multiple *FOR* statements is allowed, too. When *FOR* statements are
 nested, a cross product of the array elements returned by the individual *FOR*
 statements will be created.
 
-```
+```js
 FOR u IN users
   FOR l IN locations
     RETURN { "user" : u, "location" : l }
@@ -72,3 +72,39 @@ In this example, there are two array iterations: an outer iteration over the arr
 traversed as many times as there are elements in the outer array.  For each
 iteration, the current values of *users* and *locations* are made available for
 further processing in the variable *u* and *l*.
+
+## Options
+
+For collections and views, the `FOR` construct supports an optional `OPTIONS`
+suffix to modify behavior. The general syntax is:
+
+```js
+FOR variableName IN expression OPTIONS {option: value, ...}
+```
+
+### Index hints
+
+For collections, index hints are provided though this inline options mechanism.
+Hints can be specified in two different formats.
+
+The first format option is the simplest, just a single index name. This should
+be sufficient for many cases. Whenever there is a choice to potentially use an
+index for this `FOR` loop, the optimizer will first check if the specified index
+can be used. If so, it will use it, regardless of whether it would normally use
+a different index. If it cannot use that index, then it will fall back to its
+normal logic to select another index. If the optional `forceIndexHint: true` is
+specified, then it will not fall back, and instead generate an error.
+
+```js
+OPTIONS {indexHint: 'byName'[, forceIndexHint: <boolean>]}
+```
+
+The second is an array of index names, in order of preference. When specified
+this way, the optimizer will behave much in the same way as above, but will
+check the feasibility of each of the specified indices, in the order they are
+given, falling back to its normal logic or failing only if none of the specified
+indices are feasible.
+
+```js
+OPTIONS {indexHint: ['byName', 'byColor'][, forceIndexHint: <boolean>]}
+```
diff --git a/Documentation/Books/Manual/ReleaseNotes/NewFeatures35.md b/Documentation/Books/Manual/ReleaseNotes/NewFeatures35.md
index f097ad77e7..4b02a748f2 100644
--- a/Documentation/Books/Manual/ReleaseNotes/NewFeatures35.md
+++ b/Documentation/Books/Manual/ReleaseNotes/NewFeatures35.md
@@ -262,6 +262,14 @@ name _cannot_ be changed after index creation. No two indices on the same
 collection may share the same name, but two indices on different collections 
 may.
 
+### Index Hints in AQL
+
+Users may now take advantage of the `indexHint` inline query option to override
+the internal optimizer decision regarding which index to use to serve content
+from a given collection. The index hint works with the named indices feature
+above, making it easy to specify which index to use.
+
+
 Client tools
 ------------
 
diff --git a/arangod/Aql/Condition.cpp b/arangod/Aql/Condition.cpp
index b553ebf0a3..f6dc7dbc6c 100644
--- a/arangod/Aql/Condition.cpp
+++ b/arangod/Aql/Condition.cpp
@@ -462,12 +462,14 @@ std::pair<bool, bool> Condition::findIndexes(EnumerateCollectionNode const* node
   }
   if (_root == nullptr) {
     size_t dummy;
-    return std::make_pair<bool, bool>(false, trx->getIndexForSortCondition(collectionName, sortCondition, reference,
-                                         itemsInIndex, usedIndexes, dummy));
+    return std::make_pair<bool, bool>(
+        false, trx->getIndexForSortCondition(collectionName, sortCondition, reference, itemsInIndex,
+                                             node->hint(), usedIndexes, dummy));
   }
 
   return trx->getBestIndexHandlesForFilterCondition(collectionName, _ast, _root,
-                                                    reference, sortCondition, itemsInIndex,
+                                                    reference, sortCondition,
+                                                    itemsInIndex, node->hint(),
                                                     usedIndexes, _isSorted);
 }
 
diff --git a/arangod/Aql/ExecutionNode.cpp b/arangod/Aql/ExecutionNode.cpp
index f2f66741e1..123f58e83f 100644
--- a/arangod/Aql/ExecutionNode.cpp
+++ b/arangod/Aql/ExecutionNode.cpp
@@ -1275,7 +1275,8 @@ EnumerateCollectionNode::EnumerateCollectionNode(ExecutionPlan* plan,
     : ExecutionNode(plan, base),
       DocumentProducingNode(plan, base),
       CollectionAccessingNode(plan, base),
-      _random(base.get("random").getBoolean()) {}
+      _random(base.get("random").getBoolean()),
+      _hint(base) {}
 
 /// @brief toVelocyPack, for EnumerateCollectionNode
 void EnumerateCollectionNode::toVelocyPackHelper(VPackBuilder& builder, unsigned flags) const {
@@ -1284,6 +1285,8 @@ void EnumerateCollectionNode::toVelocyPackHelper(VPackBuilder& builder, unsigned
 
   builder.add("random", VPackValue(_random));
 
+  _hint.toVelocyPack(builder);
+
   // add outvariable and projection
   DocumentProducingNode::toVelocyPack(builder);
 
@@ -1325,7 +1328,7 @@ ExecutionNode* EnumerateCollectionNode::clone(ExecutionPlan* plan, bool withDepe
   }
 
   auto c = std::make_unique<EnumerateCollectionNode>(plan, _id, _collection,
-                                                     outVariable, _random);
+                                                     outVariable, _random, _hint);
 
   c->projections(_projections);
 
diff --git a/arangod/Aql/ExecutionNode.h b/arangod/Aql/ExecutionNode.h
index b93224fe8e..c14a089986 100644
--- a/arangod/Aql/ExecutionNode.h
+++ b/arangod/Aql/ExecutionNode.h
@@ -56,6 +56,7 @@
 #include "Aql/CostEstimate.h"
 #include "Aql/DocumentProducingNode.h"
 #include "Aql/Expression.h"
+#include "Aql/IndexHint.h"
 #include "Aql/Variable.h"
 #include "Aql/WalkerWorker.h"
 #include "Aql/types.h"
@@ -637,11 +638,12 @@ class EnumerateCollectionNode : public ExecutionNode,
   /// @brief constructor with a vocbase and a collection name
  public:
   EnumerateCollectionNode(ExecutionPlan* plan, size_t id, aql::Collection const* collection,
-                          Variable const* outVariable, bool random)
+                          Variable const* outVariable, bool random, IndexHint const& hint)
       : ExecutionNode(plan, id),
         DocumentProducingNode(outVariable),
         CollectionAccessingNode(collection),
-        _random(random) {}
+        _random(random),
+        _hint(hint) {}
 
   EnumerateCollectionNode(ExecutionPlan* plan, arangodb::velocypack::Slice const& base);
 
@@ -675,9 +677,15 @@ class EnumerateCollectionNode : public ExecutionNode,
   /// @brief enable random iteration of documents in collection
   void setRandom() { _random = true; }
 
+  /// @brief user hint regarding which index ot use
+  IndexHint const& hint() const { return _hint; }
+
  private:
   /// @brief whether or not we want random iteration
   bool _random;
+
+  /// @brief a possible hint from the user regarding which index to use
+  IndexHint _hint;
 };
 
 /// @brief class EnumerateListNode
diff --git a/arangod/Aql/ExecutionPlan.cpp b/arangod/Aql/ExecutionPlan.cpp
index 12c1fbb169..f31ae671a4 100644
--- a/arangod/Aql/ExecutionPlan.cpp
+++ b/arangod/Aql/ExecutionPlan.cpp
@@ -30,6 +30,7 @@
 #include "Aql/ExecutionNode.h"
 #include "Aql/Expression.h"
 #include "Aql/Function.h"
+#include "Aql/IndexHint.h"
 #include "Aql/ModificationNodes.h"
 #include "Aql/NodeFinder.h"
 #include "Aql/OptimizerRulesFeature.h"
@@ -842,7 +843,8 @@ ExecutionNode* ExecutionPlan::fromNodeFor(ExecutionNode* previous, AstNode const
 
   auto variable = node->getMember(0);
   auto expression = node->getMember(1);
-  // TODO: process FOR options here if we want to use them later
+  auto options = node->getMember(2);
+  IndexHint hint(options);
 
   // fetch 1st operand (out variable name)
   TRI_ASSERT(variable->type == NODE_TYPE_VARIABLE);
@@ -862,7 +864,7 @@ ExecutionNode* ExecutionPlan::fromNodeFor(ExecutionNode* previous, AstNode const
       THROW_ARANGO_EXCEPTION_MESSAGE(TRI_ERROR_INTERNAL,
                                      "no collection for EnumerateCollection");
     }
-    en = registerNode(new EnumerateCollectionNode(this, nextId(), collection, v, false));
+    en = registerNode(new EnumerateCollectionNode(this, nextId(), collection, v, false, hint));
 #ifdef USE_IRESEARCH
   } else if (expression->type == NODE_TYPE_VIEW) {
     // second operand is a view
diff --git a/arangod/Aql/IndexHint.cpp b/arangod/Aql/IndexHint.cpp
new file mode 100644
index 0000000000..9021f59d7d
--- /dev/null
+++ b/arangod/Aql/IndexHint.cpp
@@ -0,0 +1,196 @@
+////////////////////////////////////////////////////////////////////////////////
+/// DISCLAIMER
+///
+/// Copyright 2014-2019 ArangoDB GmbH, Cologne, Germany
+/// Copyright 2004-2014 triAGENS GmbH, Cologne, Germany
+///
+/// Licensed under the Apache License, Version 2.0 (the "License");
+/// you may not use this file except in compliance with the License.
+/// You may obtain a copy of the License at
+///
+///     http://www.apache.org/licenses/LICENSE-2.0
+///
+/// Unless required by applicable law or agreed to in writing, software
+/// distributed under the License is distributed on an "AS IS" BASIS,
+/// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+/// See the License for the specific language governing permissions and
+/// limitations under the License.
+///
+/// Copyright holder is ArangoDB GmbH, Cologne, Germany
+///
+/// @author Dan Larkin-York
+////////////////////////////////////////////////////////////////////////////////
+
+#include "IndexHint.h"
+
+#include <velocypack/Iterator.h>
+#include <velocypack/StringRef.h>
+
+#include "Aql/AstNode.h"
+#include "Basics/StaticStrings.h"
+#include "Basics/VelocyPackHelper.h"
+
+namespace {
+std::string const TypeIllegal("illegal");
+std::string const TypeNone("none");
+std::string const TypeSimple("simple");
+
+std::string const FieldContainer("indexHint");
+std::string const FieldForced("forced");
+std::string const FieldHint("hint");
+std::string const FieldType("type");
+
+bool extractForced(arangodb::aql::AstNode const* node) {
+  using arangodb::aql::AstNode;
+  using arangodb::aql::AstNodeType;
+  using arangodb::aql::AstNodeValueType;
+
+  bool forced = false;
+
+  if (node->type == AstNodeType::NODE_TYPE_OBJECT) {
+    for (size_t i = 0; i < node->numMembers(); i++) {
+      AstNode const* child = node->getMember(i);
+
+      if (child->type == AstNodeType::NODE_TYPE_OBJECT_ELEMENT) {
+        VPackStringRef name(child->getStringValue(), child->getStringLength());
+
+        if (name == arangodb::StaticStrings::IndexHintOptionForce) {
+          TRI_ASSERT(child->numMembers() > 0);
+          AstNode const* value = child->getMember(0);
+
+          if (value->type == AstNodeType::NODE_TYPE_VALUE &&
+              value->value.type == AstNodeValueType::VALUE_TYPE_BOOL) {
+            forced = value->value.value._bool;
+          }
+        }
+      }
+    }
+  }
+
+  return forced;
+}
+
+arangodb::aql::IndexHint::HintType fromTypeName(std::string const& typeName) {
+  if (::TypeSimple == typeName) {
+    return arangodb::aql::IndexHint::HintType::Simple;
+  } else if (::TypeNone == typeName) {
+    return arangodb::aql::IndexHint::HintType::None;
+  }
+
+  return arangodb::aql::IndexHint::HintType::Illegal;
+}
+}  // namespace
+
+namespace arangodb {
+namespace aql {
+
+IndexHint::IndexHint() : _type{HintType::None}, _forced{false} {}
+
+IndexHint::IndexHint(AstNode const* node)
+    : _type{HintType::None}, _forced{::extractForced(node)} {
+  if (node->type == AstNodeType::NODE_TYPE_OBJECT) {
+    for (size_t i = 0; i < node->numMembers(); i++) {
+      AstNode const* child = node->getMember(i);
+
+      if (child->type == AstNodeType::NODE_TYPE_OBJECT_ELEMENT) {
+        VPackStringRef name(child->getStringValue(), child->getStringLength());
+
+        if (name == StaticStrings::IndexHintOption) {
+          TRI_ASSERT(child->numMembers() > 0);
+          AstNode const* value = child->getMember(0);
+
+          if (value->type == AstNodeType::NODE_TYPE_VALUE &&
+              value->value.type == AstNodeValueType::VALUE_TYPE_STRING) {
+            _type = HintType::Simple;
+            _hint.simple.emplace_back(value->getStringValue(), value->getStringLength());
+          }
+
+          if (value->type == AstNodeType::NODE_TYPE_ARRAY) {
+            _type = HintType::Simple;
+            for (size_t j = 0; j < value->numMembers(); j++) {
+              AstNode const* member = value->getMember(j);
+              if (member->type == AstNodeType::NODE_TYPE_VALUE &&
+                  member->value.type == AstNodeValueType::VALUE_TYPE_STRING) {
+                _hint.simple.emplace_back(member->getStringValue(),
+                                          member->getStringLength());
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+IndexHint::IndexHint(VPackSlice const& slice)
+    : _type{::fromTypeName(
+          basics::VelocyPackHelper::getStringValue(slice.get(::FieldContainer),
+                                                   ::FieldType, ""))},
+      _forced{basics::VelocyPackHelper::getBooleanValue(slice.get(::FieldContainer),
+                                                        ::FieldForced, false)} {
+  if (_type != HintType::Illegal && _type != HintType::None) {
+    VPackSlice container = slice.get(::FieldContainer);
+    TRI_ASSERT(container.isObject());
+
+    if (_type == HintType::Simple) {
+      VPackSlice hintSlice = container.get(::FieldHint);
+      TRI_ASSERT(hintSlice.isArray());
+      for (VPackSlice index : VPackArrayIterator(hintSlice)) {
+        TRI_ASSERT(index.isString());
+        _hint.simple.emplace_back(index.copyString());
+      }
+    }
+  }
+}
+
+IndexHint::HintType IndexHint::type() const { return _type; }
+
+bool IndexHint::isForced() const { return _forced; }
+
+std::vector<std::string> const& IndexHint::hint() const {
+  TRI_ASSERT(_type == HintType::Simple);
+  return _hint.simple;
+}
+
+std::string IndexHint::typeName() const {
+  switch (_type) {
+    case HintType::Illegal:
+      return ::TypeIllegal;
+    case HintType::None:
+      return ::TypeNone;
+    case HintType::Simple:
+      return ::TypeSimple;
+  }
+
+  return ::TypeIllegal;
+}
+
+void IndexHint::toVelocyPack(VPackBuilder& builder) const {
+  TRI_ASSERT(builder.isOpenObject());
+  VPackObjectBuilder guard(&builder, ::FieldContainer);
+  builder.add(::FieldForced, VPackValue(_forced));
+  builder.add(::FieldType, VPackValue(typeName()));
+  if (_type == HintType::Simple) {
+    VPackArrayBuilder hintGuard(&builder, ::FieldHint);
+    for (std::string const& index : _hint.simple) {
+      builder.add(VPackValue(index));
+    }
+  }
+}
+
+std::string IndexHint::toString() const {
+  VPackBuilder builder;
+  {
+    VPackObjectBuilder guard(&builder);
+    toVelocyPack(builder);
+  }
+  return builder.slice().toJson();
+}
+
+std::ostream& operator<<(std::ostream& stream, arangodb::aql::IndexHint const& hint) {
+  stream << hint.toString();
+  return stream;
+}
+
+}  // namespace aql
+}  // namespace arangodb
diff --git a/arangod/Aql/IndexHint.h b/arangod/Aql/IndexHint.h
new file mode 100644
index 0000000000..69acce300b
--- /dev/null
+++ b/arangod/Aql/IndexHint.h
@@ -0,0 +1,74 @@
+////////////////////////////////////////////////////////////////////////////////
+/// DISCLAIMER
+///
+/// Copyright 2014-2019 ArangoDB GmbH, Cologne, Germany
+/// Copyright 2004-2014 triAGENS GmbH, Cologne, Germany
+///
+/// Licensed under the Apache License, Version 2.0 (the "License");
+/// you may not use this file except in compliance with the License.
+/// You may obtain a copy of the License at
+///
+///     http://www.apache.org/licenses/LICENSE-2.0
+///
+/// Unless required by applicable law or agreed to in writing, software
+/// distributed under the License is distributed on an "AS IS" BASIS,
+/// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+/// See the License for the specific language governing permissions and
+/// limitations under the License.
+///
+/// Copyright holder is ArangoDB GmbH, Cologne, Germany
+///
+/// @author Dan Larkin-York
+////////////////////////////////////////////////////////////////////////////////
+
+#ifndef ARANGOD_AQL_INDEX_HINT_H
+#define ARANGOD_AQL_INDEX_HINT_H 1
+
+#include <iostream>
+
+#include <velocypack/Builder.h>
+#include <velocypack/Slice.h>
+#include <velocypack/velocypack-aliases.h>
+
+#include "Aql/AstNode.h"
+
+namespace arangodb {
+namespace aql {
+
+/// @brief container for index hint information
+class IndexHint {
+ public:
+  enum HintType : uint8_t { Illegal, None, Simple };
+
+ public:
+  explicit IndexHint();
+  explicit IndexHint(AstNode const* node);
+  explicit IndexHint(VPackSlice const& slice);
+
+ public:
+  HintType type() const;
+  bool isForced() const;
+  std::vector<std::string> const& hint() const;
+
+  void toVelocyPack(VPackBuilder& builder) const;
+  std::string typeName() const;
+  std::string toString() const;
+
+ private:
+  HintType _type;
+  bool const _forced;
+
+  // actual hint is a recursive structure, with the data type determined by the
+  // _type above; in the case of a nested IndexHint, the value of isForced() is
+  // inherited
+  struct HintData {
+    std::vector<std::string> simple;
+  } _hint;
+};
+
+std::ostream& operator<<(std::ostream& stream, arangodb::aql::IndexHint const& hint);
+
+}  // namespace aql
+}  // namespace arangodb
+
+#endif
diff --git a/arangod/Aql/OptimizerRules.cpp b/arangod/Aql/OptimizerRules.cpp
index 15446c7eaa..4613922e4e 100644
--- a/arangod/Aql/OptimizerRules.cpp
+++ b/arangod/Aql/OptimizerRules.cpp
@@ -2880,11 +2880,10 @@ struct SortToIndexNode final : public WalkerWorker<ExecutionNode> {
       std::vector<transaction::Methods::IndexHandle> usedIndexes;
       auto trx = _plan->getAst()->query()->trx();
       size_t coveredAttributes = 0;
-      bool canBeUsed = 
-          trx->getIndexForSortCondition(enumerateCollectionNode->collection()->name(),
-                                        &sortCondition, outVariable,
-                                        enumerateCollectionNode->collection()->count(trx),
-                                        usedIndexes, coveredAttributes);
+      bool canBeUsed = trx->getIndexForSortCondition(
+          enumerateCollectionNode->collection()->name(), &sortCondition,
+          outVariable, enumerateCollectionNode->collection()->count(trx),
+          enumerateCollectionNode->hint(), usedIndexes, coveredAttributes);
       if (canBeUsed) {
         // If this bit is set, then usedIndexes has length exactly one
         // and contains the best index found.
diff --git a/arangod/Aql/OptimizerRulesReplaceFunctions.cpp b/arangod/Aql/OptimizerRulesReplaceFunctions.cpp
index cf912da7f9..33a275a5fe 100644
--- a/arangod/Aql/OptimizerRulesReplaceFunctions.cpp
+++ b/arangod/Aql/OptimizerRulesReplaceFunctions.cpp
@@ -24,6 +24,7 @@
 #include "Aql/ExecutionNode.h"
 #include "Aql/ExecutionPlan.h"
 #include "Aql/Function.h"
+#include "Aql/IndexHint.h"
 #include "Aql/IndexNode.h"
 #include "Aql/Optimizer.h"
 #include "Aql/Query.h"
@@ -266,7 +267,7 @@ AstNode* replaceNearOrWithin(AstNode* funAstNode, ExecutionNode* calcNode,
   ExecutionNode* eEnumerate = plan->registerNode(
       // link output of index with the return node
       new EnumerateCollectionNode(plan, plan->nextId(), aqlCollection,
-                                  enumerateOutVariable, false));
+                                  enumerateOutVariable, false, IndexHint()));
 
   //// build sort condition - DISTANCE(d.lat, d.long, param.lat, param.lon)
   auto* docRef = ast->createNodeReference(enumerateOutVariable);
diff --git a/arangod/CMakeLists.txt b/arangod/CMakeLists.txt
index b68c531f23..28204b87f3 100644
--- a/arangod/CMakeLists.txt
+++ b/arangod/CMakeLists.txt
@@ -227,6 +227,7 @@ SET(ARANGOD_SOURCES
   Aql/InAndOutRowExpressionContext.cpp
   Aql/IdExecutor.cpp
   Aql/IndexExecutor.cpp
+  Aql/IndexHint.cpp
   Aql/IndexNode.cpp
   Aql/InputAqlItemRow.cpp
   Aql/LimitExecutor.cpp
diff --git a/arangod/Cluster/ClusterMethods.cpp b/arangod/Cluster/ClusterMethods.cpp
index 3b3f2e9d02..f3ff51aaae 100644
--- a/arangod/Cluster/ClusterMethods.cpp
+++ b/arangod/Cluster/ClusterMethods.cpp
@@ -31,6 +31,7 @@
 #include "Basics/tri-strings.h"
 #include "Cluster/ClusterComm.h"
 #include "Cluster/ClusterInfo.h"
+#include "ClusterMethods.h"
 #include "Graph/Traverser.h"
 #include "Indexes/Index.h"
 #include "RestServer/TtlFeature.h"
diff --git a/arangod/Graph/BaseOptions.cpp b/arangod/Graph/BaseOptions.cpp
index 4b8170a076..5ab8689d71 100644
--- a/arangod/Graph/BaseOptions.cpp
+++ b/arangod/Graph/BaseOptions.cpp
@@ -248,9 +248,10 @@ void BaseOptions::injectLookupInfoInList(std::vector<LookupInfo>& list,
                                          aql::AstNode* condition) {
   LookupInfo info;
   info.indexCondition = condition->clone(plan->getAst());
-  bool res = _trx->getBestIndexHandleForFilterCondition(collectionName,
-                                                        info.indexCondition, _tmpVar,
-                                                        1000, info.idxHandles[0]);
+  bool res =
+      _trx->getBestIndexHandleForFilterCondition(collectionName, info.indexCondition,
+                                                 _tmpVar, 1000, aql::IndexHint(),
+                                                 info.idxHandles[0]);
   // Right now we have an enforced edge index which should always fit.
   if (!res) {
     THROW_ARANGO_EXCEPTION_MESSAGE(TRI_ERROR_INTERNAL,
diff --git a/arangod/Graph/EdgeCollectionInfo.cpp b/arangod/Graph/EdgeCollectionInfo.cpp
index daa5b25b2d..cce93a8bdf 100644
--- a/arangod/Graph/EdgeCollectionInfo.cpp
+++ b/arangod/Graph/EdgeCollectionInfo.cpp
@@ -46,21 +46,25 @@ EdgeCollectionInfo::EdgeCollectionInfo(transaction::Methods* trx,
   auto var = _searchBuilder.getVariable();
   if (_dir == TRI_EDGE_OUT) {
     auto cond = _searchBuilder.getOutboundCondition();
-    bool worked = _trx->getBestIndexHandleForFilterCondition(_collectionName, cond,
-                                                             var, 1000, _forwardIndexId);
+    bool worked =
+        _trx->getBestIndexHandleForFilterCondition(_collectionName, cond, var, 1000,
+                                                   aql::IndexHint(), _forwardIndexId);
     TRI_ASSERT(worked);  // We always have an edge Index
     cond = _searchBuilder.getInboundCondition();
     worked = _trx->getBestIndexHandleForFilterCondition(_collectionName, cond, var,
-                                                        1000, _backwardIndexId);
+                                                        1000, aql::IndexHint(),
+                                                        _backwardIndexId);
     TRI_ASSERT(worked);  // We always have an edge Index
   } else {
     auto cond = _searchBuilder.getInboundCondition();
-    bool worked = _trx->getBestIndexHandleForFilterCondition(_collectionName, cond,
-                                                             var, 1000, _forwardIndexId);
+    bool worked =
+        _trx->getBestIndexHandleForFilterCondition(_collectionName, cond, var, 1000,
+                                                   aql::IndexHint(), _forwardIndexId);
     TRI_ASSERT(worked);  // We always have an edge Index
     cond = _searchBuilder.getOutboundCondition();
     worked = _trx->getBestIndexHandleForFilterCondition(_collectionName, cond, var,
-                                                        1000, _backwardIndexId);
+                                                        1000, aql::IndexHint(),
+                                                        _backwardIndexId);
     TRI_ASSERT(worked);  // We always have an edge Index
   }
 }
diff --git a/arangod/RestHandler/RestEdgesHandler.cpp b/arangod/RestHandler/RestEdgesHandler.cpp
index cfbdf2501a..213b19b3e7 100644
--- a/arangod/RestHandler/RestEdgesHandler.cpp
+++ b/arangod/RestHandler/RestEdgesHandler.cpp
@@ -67,8 +67,9 @@ void RestEdgesHandler::readCursor(aql::AstNode* condition, aql::Variable const*
                                   SingleCollectionTransaction& trx,
                                   std::function<void(LocalDocumentId const&)> const& cb) {
   transaction::Methods::IndexHandle indexId;
-  bool foundIdx = trx.getBestIndexHandleForFilterCondition(collectionName, condition,
-                                                           var, 1000, indexId);
+  bool foundIdx =
+      trx.getBestIndexHandleForFilterCondition(collectionName, condition, var,
+                                               1000, aql::IndexHint(), indexId);
   if (!foundIdx) {
     // Right now we enforce an edge index that can exactly! work on this
     // condition. So it is impossible to not find an index.
diff --git a/arangod/Transaction/Methods.cpp b/arangod/Transaction/Methods.cpp
index b533af5f44..4ebf649ad3 100644
--- a/arangod/Transaction/Methods.cpp
+++ b/arangod/Transaction/Methods.cpp
@@ -576,14 +576,16 @@ std::pair<bool, bool> transaction::Methods::findIndexHandleForAndNode(
     std::vector<std::shared_ptr<Index>> const& indexes,
     arangodb::aql::AstNode* node, arangodb::aql::Variable const* reference,
     arangodb::aql::SortCondition const* sortCondition, size_t itemsInCollection,
-    std::vector<transaction::Methods::IndexHandle>& usedIndexes,
+    aql::IndexHint const& hint, std::vector<transaction::Methods::IndexHandle>& usedIndexes,
     arangodb::aql::AstNode*& specializedCondition, bool& isSparse) const {
   std::shared_ptr<Index> bestIndex;
   double bestCost = 0.0;
   bool bestSupportsFilter = false;
   bool bestSupportsSort = false;
 
-  for (auto const& idx : indexes) {
+  auto considerIndex = [&bestIndex, &bestCost, &bestSupportsFilter, &bestSupportsSort,
+                        &indexes, node, reference, itemsInCollection,
+                        sortCondition](std::shared_ptr<Index> const& idx) -> void {
     double filterCost = 0.0;
     double sortCost = 0.0;
     size_t itemsInIndex = itemsInCollection;
@@ -636,7 +638,7 @@ std::pair<bool, bool> transaction::Methods::findIndexHandleForAndNode(
     }
 
     if (!supportsFilter && !supportsSort) {
-      continue;
+      return;
     }
 
     double totalCost = filterCost;
@@ -666,6 +668,38 @@ std::pair<bool, bool> transaction::Methods::findIndexHandleForAndNode(
       bestSupportsFilter = supportsFilter;
       bestSupportsSort = supportsSort;
     }
+  };
+
+  if (hint.type() == aql::IndexHint::HintType::Simple) {
+    std::vector<std::string> const& hintedIndices = hint.hint();
+    for (std::string const& hinted : hintedIndices) {
+      std::shared_ptr<Index> matched;
+      for (std::shared_ptr<Index> const& idx : indexes) {
+        if (idx->name() == hinted) {
+          matched = idx;
+          break;
+        }
+      }
+
+      if (matched != nullptr) {
+        considerIndex(matched);
+        if (bestIndex != nullptr) {
+          break;
+        }
+      }
+    }
+
+    if (hint.isForced() && bestIndex == nullptr) {
+      THROW_ARANGO_EXCEPTION_MESSAGE(
+          TRI_ERROR_QUERY_FORCED_INDEX_HINT_UNUSABLE,
+          "could not use index hint to serve query; " + hint.toString());
+    }
+  }
+
+  if (bestIndex == nullptr) {
+    for (auto const& idx : indexes) {
+      considerIndex(idx);
+    }
   }
 
   if (bestIndex == nullptr) {
@@ -681,13 +715,14 @@ std::pair<bool, bool> transaction::Methods::findIndexHandleForAndNode(
 }
 
 bool transaction::Methods::findIndexHandleForAndNode(
-    std::vector<std::shared_ptr<Index>> const& indexes,
-    arangodb::aql::AstNode*& node, arangodb::aql::Variable const* reference,
-    size_t itemsInCollection, transaction::Methods::IndexHandle& usedIndex) const {
+    std::vector<std::shared_ptr<Index>> const& indexes, arangodb::aql::AstNode*& node,
+    arangodb::aql::Variable const* reference, size_t itemsInCollection,
+    aql::IndexHint const& hint, transaction::Methods::IndexHandle& usedIndex) const {
   std::shared_ptr<Index> bestIndex;
   double bestCost = 0.0;
 
-  for (auto const& idx : indexes) {
+  auto considerIndex = [&bestIndex, &bestCost, itemsInCollection, &indexes, &node,
+                        reference](std::shared_ptr<Index> const& idx) -> void {
     size_t itemsInIndex = itemsInCollection;
 
     // check if the index supports the filter expression
@@ -709,7 +744,7 @@ bool transaction::Methods::findIndexHandleForAndNode(
         << ", node: " << node;
 
     if (!supportsFilter) {
-      continue;
+      return;
     }
 
     // index supports the filter condition
@@ -721,6 +756,38 @@ bool transaction::Methods::findIndexHandleForAndNode(
       bestIndex = idx;
       bestCost = estimatedCost;
     }
+  };
+
+  if (hint.type() == aql::IndexHint::HintType::Simple) {
+    std::vector<std::string> const& hintedIndices = hint.hint();
+    for (std::string const& hinted : hintedIndices) {
+      std::shared_ptr<Index> matched;
+      for (std::shared_ptr<Index> const& idx : indexes) {
+        if (idx->name() == hinted) {
+          matched = idx;
+          break;
+        }
+      }
+
+      if (matched != nullptr) {
+        considerIndex(matched);
+        if (bestIndex != nullptr) {
+          break;
+        }
+      }
+    }
+
+    if (hint.isForced() && bestIndex == nullptr) {
+      THROW_ARANGO_EXCEPTION_MESSAGE(
+          TRI_ERROR_QUERY_FORCED_INDEX_HINT_UNUSABLE,
+          "could not use index hint to serve query; " + hint.toString());
+    }
+  }
+
+  if (bestIndex == nullptr) {
+    for (auto const& idx : indexes) {
+      considerIndex(idx);
+    }
   }
 
   if (bestIndex == nullptr) {
@@ -2755,7 +2822,7 @@ std::pair<bool, bool> transaction::Methods::getBestIndexHandlesForFilterConditio
     std::string const& collectionName, arangodb::aql::Ast* ast,
     arangodb::aql::AstNode* root, arangodb::aql::Variable const* reference,
     arangodb::aql::SortCondition const* sortCondition, size_t itemsInCollection,
-    std::vector<IndexHandle>& usedIndexes, bool& isSorted) {
+    aql::IndexHint const& hint, std::vector<IndexHandle>& usedIndexes, bool& isSorted) {
   // We can only start after DNF transformation
   TRI_ASSERT(root->type == arangodb::aql::AstNodeType::NODE_TYPE_OPERATOR_NARY_OR);
   auto indexes = indexesForCollection(collectionName);
@@ -2771,7 +2838,7 @@ std::pair<bool, bool> transaction::Methods::getBestIndexHandlesForFilterConditio
     auto node = root->getMemberUnchecked(i);
     arangodb::aql::AstNode* specializedCondition = nullptr;
     auto canUseIndex = findIndexHandleForAndNode(indexes, node, reference, sortCondition,
-                                                 itemsInCollection, usedIndexes,
+                                                 itemsInCollection, hint, usedIndexes,
                                                  specializedCondition, isSparse);
 
     if (canUseIndex.second && !canUseIndex.first) {
@@ -2817,7 +2884,7 @@ std::pair<bool, bool> transaction::Methods::getBestIndexHandlesForFilterConditio
 bool transaction::Methods::getBestIndexHandleForFilterCondition(
     std::string const& collectionName, arangodb::aql::AstNode*& node,
     arangodb::aql::Variable const* reference, size_t itemsInCollection,
-    IndexHandle& usedIndex) {
+    aql::IndexHint const& hint, IndexHandle& usedIndex) {
   // We can only start after DNF transformation and only a single AND
   TRI_ASSERT(node->type == arangodb::aql::AstNodeType::NODE_TYPE_OPERATOR_NARY_AND);
   if (node->numMembers() == 0) {
@@ -2829,7 +2896,8 @@ bool transaction::Methods::getBestIndexHandleForFilterCondition(
 
   // Const cast is save here. Giving computeSpecialization == false
   // Makes sure node is NOT modified.
-  return findIndexHandleForAndNode(indexes, node, reference, itemsInCollection, usedIndex);
+  return findIndexHandleForAndNode(indexes, node, reference, itemsInCollection,
+                                   hint, usedIndex);
 }
 
 /// @brief Checks if the index supports the filter condition.
@@ -2872,20 +2940,20 @@ std::vector<std::vector<arangodb::basics::AttributeName>> transaction::Methods::
 bool transaction::Methods::getIndexForSortCondition(
     std::string const& collectionName, arangodb::aql::SortCondition const* sortCondition,
     arangodb::aql::Variable const* reference, size_t itemsInIndex,
-    std::vector<IndexHandle>& usedIndexes, size_t& coveredAttributes) {
+    aql::IndexHint const& hint, std::vector<IndexHandle>& usedIndexes,
+    size_t& coveredAttributes) {
   // We do not have a condition. But we have a sort!
   if (!sortCondition->isEmpty() && sortCondition->isOnlyAttributeAccess() &&
       sortCondition->isUnidirectional()) {
     double bestCost = 0.0;
     std::shared_ptr<Index> bestIndex;
 
-    auto indexes = indexesForCollection(collectionName);
-
-    for (auto const& idx : indexes) {
+    auto considerIndex = [reference, sortCondition, itemsInIndex, &bestCost, &bestIndex,
+                          &coveredAttributes](std::shared_ptr<Index> const& idx) -> void {
       if (idx->sparse()) {
         // a sparse index may exclude some documents, so it can't be used to
         // get a sorted view of the ENTIRE collection
-        continue;
+        return;
       }
       double sortCost = 0.0;
       size_t covered = 0;
@@ -2897,6 +2965,40 @@ bool transaction::Methods::getIndexForSortCondition(
           coveredAttributes = covered;
         }
       }
+    };
+
+    auto indexes = indexesForCollection(collectionName);
+
+    if (hint.type() == aql::IndexHint::HintType::Simple) {
+      std::vector<std::string> const& hintedIndices = hint.hint();
+      for (std::string const& hinted : hintedIndices) {
+        std::shared_ptr<Index> matched;
+        for (std::shared_ptr<Index> const& idx : indexes) {
+          if (idx->name() == hinted) {
+            matched = idx;
+            break;
+          }
+        }
+
+        if (matched != nullptr) {
+          considerIndex(matched);
+          if (bestIndex != nullptr) {
+            break;
+          }
+        }
+      }
+
+      if (hint.isForced() && bestIndex == nullptr) {
+        THROW_ARANGO_EXCEPTION_MESSAGE(
+            TRI_ERROR_QUERY_FORCED_INDEX_HINT_UNUSABLE,
+            "could not use index hint to serve query; " + hint.toString());
+      }
+    }
+
+    if (bestIndex == nullptr) {
+      for (auto const& idx : indexes) {
+        considerIndex(idx);
+      }
     }
 
     if (bestIndex != nullptr) {
diff --git a/arangod/Transaction/Methods.h b/arangod/Transaction/Methods.h
index fea5f475bc..fa258e48f9 100644
--- a/arangod/Transaction/Methods.h
+++ b/arangod/Transaction/Methods.h
@@ -24,6 +24,7 @@
 #ifndef ARANGOD_TRANSACTION_METHODS_H
 #define ARANGOD_TRANSACTION_METHODS_H 1
 
+#include "Aql/IndexHint.h"
 #include "Basics/Common.h"
 #include "Basics/Exceptions.h"
 #include "Basics/Result.h"
@@ -331,7 +332,7 @@ class Methods {
   ENTERPRISE_VIRT std::pair<bool, bool> getBestIndexHandlesForFilterCondition(
       std::string const&, arangodb::aql::Ast*, arangodb::aql::AstNode*,
       arangodb::aql::Variable const*, arangodb::aql::SortCondition const*,
-      size_t, std::vector<IndexHandle>&, bool&);
+      size_t, aql::IndexHint const&, std::vector<IndexHandle>&, bool&);
 
   /// @brief Gets the best fitting index for one specific condition.
   ///        Difference to IndexHandles: Condition is only one NARY_AND
@@ -339,10 +340,9 @@ class Methods {
   ///        Returns false if no index could be found.
   ///        If it returned true, the AstNode contains the specialized condition
 
-  ENTERPRISE_VIRT bool getBestIndexHandleForFilterCondition(std::string const&,
-                                                            arangodb::aql::AstNode*&,
-                                                            arangodb::aql::Variable const*,
-                                                            size_t, IndexHandle&);
+  ENTERPRISE_VIRT bool getBestIndexHandleForFilterCondition(
+      std::string const&, arangodb::aql::AstNode*&,
+      arangodb::aql::Variable const*, size_t, aql::IndexHint const&, IndexHandle&);
 
   /// @brief Checks if the index supports the filter condition.
   /// note: the caller must have read-locked the underlying collection when
@@ -362,7 +362,8 @@ class Methods {
   ENTERPRISE_VIRT bool getIndexForSortCondition(std::string const&,
                                                 arangodb::aql::SortCondition const*,
                                                 arangodb::aql::Variable const*,
-                                                size_t, std::vector<IndexHandle>&,
+                                                size_t, aql::IndexHint const&,
+                                                std::vector<IndexHandle>&,
                                                 size_t& coveredAttributes);
 
   /// @brief factory for IndexIterator objects from AQL
@@ -562,13 +563,14 @@ class Methods {
       std::vector<std::shared_ptr<Index>> const& indexes,
       arangodb::aql::AstNode* node, arangodb::aql::Variable const* reference,
       arangodb::aql::SortCondition const* sortCondition, size_t itemsInCollection,
-      std::vector<transaction::Methods::IndexHandle>& usedIndexes,
+      aql::IndexHint const& hint, std::vector<transaction::Methods::IndexHandle>& usedIndexes,
       arangodb::aql::AstNode*& specializedCondition, bool& isSparse) const;
 
   /// @brief findIndexHandleForAndNode, Shorthand which does not support Sort
   bool findIndexHandleForAndNode(std::vector<std::shared_ptr<Index>> const& indexes,
                                  arangodb::aql::AstNode*& node,
-                                 arangodb::aql::Variable const* reference, size_t itemsInCollection,
+                                 arangodb::aql::Variable const* reference,
+                                 size_t itemsInCollection, aql::IndexHint const& hint,
                                  transaction::Methods::IndexHandle& usedIndex) const;
 
   /// @brief Get one index by id for a collection name, coordinator case
diff --git a/js/common/bootstrap/errors.js b/js/common/bootstrap/errors.js
index 45d0b3ba3d..c0601dc4dd 100644
--- a/js/common/bootstrap/errors.js
+++ b/js/common/bootstrap/errors.js
@@ -215,6 +215,7 @@
     "ERROR_QUERY_INVALID_AGGREGATE_EXPRESSION" : { "code" : 1574, "message" : "invalid aggregate expression" },
     "ERROR_QUERY_COMPILE_TIME_OPTIONS" : { "code" : 1575, "message" : "query options must be readable at query compile time" },
     "ERROR_QUERY_EXCEPTION_OPTIONS" : { "code" : 1576, "message" : "query options expected" },
+    "ERROR_QUERY_FORCED_INDEX_HINT_UNUSABLE" : { "code" : 1577, "message" : "could not use forced index hint" },
     "ERROR_QUERY_DISALLOWED_DYNAMIC_CALL" : { "code" : 1578, "message" : "disallowed dynamic call to '%s'" },
     "ERROR_QUERY_ACCESS_AFTER_MODIFICATION" : { "code" : 1579, "message" : "access after data-modification by %s" },
     "ERROR_QUERY_FUNCTION_INVALID_NAME" : { "code" : 1580, "message" : "invalid user function name" },
diff --git a/js/common/modules/@arangodb/aql/explainer.js b/js/common/modules/@arangodb/aql/explainer.js
index 3300217d13..2805a2afa2 100644
--- a/js/common/modules/@arangodb/aql/explainer.js
+++ b/js/common/modules/@arangodb/aql/explainer.js
@@ -294,6 +294,7 @@ function printIndexes(indexes) {
     var maxCollectionLen = String('Collection').length;
     var maxUniqueLen = String('Unique').length;
     var maxSparseLen = String('Sparse').length;
+    var maxNameLen = String('Name').length;
     var maxTypeLen = String('Type').length;
     var maxSelectivityLen = String('Selectivity').length;
     var maxFieldsLen = String('Fields').length;
@@ -302,6 +303,10 @@ function printIndexes(indexes) {
       if (l > maxIdLen) {
         maxIdLen = l;
       }
+      l = index.name.length;
+      if (l > maxNameLen) {
+        maxNameLen = l;
+      }
       l = index.type.length;
       if (l > maxTypeLen) {
         maxTypeLen = l;
@@ -316,6 +321,7 @@ function printIndexes(indexes) {
       }
     });
     var line = ' ' + pad(1 + maxIdLen - String('By').length) + header('By') + '   ' +
+      header('Name') + pad(1 + maxNameLen - 'Name'.length) + '   ' +
       header('Type') + pad(1 + maxTypeLen - 'Type'.length) + '   ' +
       header('Collection') + pad(1 + maxCollectionLen - 'Collection'.length) + '   ' +
       header('Unique') + pad(1 + maxUniqueLen - 'Unique'.length) + '   ' +
@@ -344,6 +350,7 @@ function printIndexes(indexes) {
       );
       line = ' ' +
         pad(1 + maxIdLen - String(indexes[i].node).length) + variable(String(indexes[i].node)) + '   ' +
+        collection(indexes[i].name) + pad(1 + maxNameLen - indexes[i].name.length) + '   ' +
         keyword(indexes[i].type) + pad(1 + maxTypeLen - indexes[i].type.length) + '   ' +
         collection(indexes[i].collection) + pad(1 + maxCollectionLen - indexes[i].collection.length) + '   ' +
         value(uniqueness) + pad(1 + maxUniqueLen - uniqueness.length) + '   ' +
diff --git a/lib/Basics/StaticStrings.cpp b/lib/Basics/StaticStrings.cpp
index f552212efe..53123954a0 100644
--- a/lib/Basics/StaticStrings.cpp
+++ b/lib/Basics/StaticStrings.cpp
@@ -107,6 +107,17 @@ std::string const StaticStrings::IndexNameEdgeTo("edge_to");
 std::string const StaticStrings::IndexNameInaccessible("inaccessible");
 std::string const StaticStrings::IndexNamePrimary("primary");
 
+// index hint strings
+std::string const StaticStrings::IndexHintAny("any");
+std::string const StaticStrings::IndexHintCollection("collection");
+std::string const StaticStrings::IndexHintHint("hint");
+std::string const StaticStrings::IndexHintDepth("depth");
+std::string const StaticStrings::IndexHintInbound("inbound");
+std::string const StaticStrings::IndexHintOption("indexHint");
+std::string const StaticStrings::IndexHintOptionForce("forceIndexHint");
+std::string const StaticStrings::IndexHintOutbound("outbound");
+std::string const StaticStrings::IndexHintWildcard("*");
+
 // HTTP headers
 std::string const StaticStrings::Accept("accept");
 std::string const StaticStrings::AcceptEncoding("accept-encoding");
diff --git a/lib/Basics/StaticStrings.h b/lib/Basics/StaticStrings.h
index 23b4aa23a2..1fbff11f2e 100644
--- a/lib/Basics/StaticStrings.h
+++ b/lib/Basics/StaticStrings.h
@@ -106,6 +106,17 @@ class StaticStrings {
   static std::string const IndexNameInaccessible;
   static std::string const IndexNamePrimary;
 
+  // index hint strings
+  static std::string const IndexHintAny;
+  static std::string const IndexHintCollection;
+  static std::string const IndexHintHint;
+  static std::string const IndexHintDepth;
+  static std::string const IndexHintInbound;
+  static std::string const IndexHintOption;
+  static std::string const IndexHintOptionForce;
+  static std::string const IndexHintOutbound;
+  static std::string const IndexHintWildcard;
+
   // HTTP headers
   static std::string const Accept;
   static std::string const AcceptEncoding;
diff --git a/lib/Basics/errors.dat b/lib/Basics/errors.dat
index e84dd93e8b..3f7b013d5c 100755
--- a/lib/Basics/errors.dat
+++ b/lib/Basics/errors.dat
@@ -258,6 +258,7 @@ ERROR_QUERY_MULTI_MODIFY,1573,"multi-modify query","Will be raised when an AQL q
 ERROR_QUERY_INVALID_AGGREGATE_EXPRESSION,1574,"invalid aggregate expression","Will be raised when an AQL query contains an invalid aggregate expression."
 ERROR_QUERY_COMPILE_TIME_OPTIONS,1575,"query options must be readable at query compile time","Will be raised when an AQL data-modification query contains options that cannot be figured out at query compile time."
 ERROR_QUERY_EXCEPTION_OPTIONS,1576,"query options expected","Will be raised when an AQL data-modification query contains an invalid options specification."
+ERROR_QUERY_FORCED_INDEX_HINT_UNUSABLE,1577,"could not use forced index hint","Will be raised when forceIndexHint is specified, and the hint cannot be used to serve the query."
 ERROR_QUERY_DISALLOWED_DYNAMIC_CALL,1578,"disallowed dynamic call to '%s'","Will be raised when a dynamic function call is made to a function that cannot be called dynamically."
 ERROR_QUERY_ACCESS_AFTER_MODIFICATION,1579,"access after data-modification by %s","Will be raised when collection data are accessed after a data-modification operation."
 
diff --git a/lib/Basics/voc-errors.cpp b/lib/Basics/voc-errors.cpp
index d143f32736..ade6748b2a 100644
--- a/lib/Basics/voc-errors.cpp
+++ b/lib/Basics/voc-errors.cpp
@@ -214,6 +214,7 @@ void TRI_InitializeErrorMessages() {
   REG_ERROR(ERROR_QUERY_INVALID_AGGREGATE_EXPRESSION, "invalid aggregate expression");
   REG_ERROR(ERROR_QUERY_COMPILE_TIME_OPTIONS, "query options must be readable at query compile time");
   REG_ERROR(ERROR_QUERY_EXCEPTION_OPTIONS, "query options expected");
+  REG_ERROR(ERROR_QUERY_FORCED_INDEX_HINT_UNUSABLE, "could not use forced index hint");
   REG_ERROR(ERROR_QUERY_DISALLOWED_DYNAMIC_CALL, "disallowed dynamic call to '%s'");
   REG_ERROR(ERROR_QUERY_ACCESS_AFTER_MODIFICATION, "access after data-modification by %s");
   REG_ERROR(ERROR_QUERY_FUNCTION_INVALID_NAME, "invalid user function name");
diff --git a/lib/Basics/voc-errors.h b/lib/Basics/voc-errors.h
index fae732a449..816d519ecc 100644
--- a/lib/Basics/voc-errors.h
+++ b/lib/Basics/voc-errors.h
@@ -1148,6 +1148,12 @@ constexpr int TRI_ERROR_QUERY_COMPILE_TIME_OPTIONS
 /// options specification.
 constexpr int TRI_ERROR_QUERY_EXCEPTION_OPTIONS                                 = 1576;
 
+/// 1577: ERROR_QUERY_FORCED_INDEX_HINT_UNUSABLE
+/// "could not use forced index hint"
+/// Will be raised when forceIndexHint is specified, and the hint cannot be
+/// used to serve the query.
+constexpr int TRI_ERROR_QUERY_FORCED_INDEX_HINT_UNUSABLE                        = 1577;
+
 /// 1578: ERROR_QUERY_DISALLOWED_DYNAMIC_CALL
 /// "disallowed dynamic call to '%s'"
 /// Will be raised when a dynamic function call is made to a function that
diff --git a/tests/js/server/aql/aql-index-hints.js b/tests/js/server/aql/aql-index-hints.js
new file mode 100644
index 0000000000..5122a5ab97
--- /dev/null
+++ b/tests/js/server/aql/aql-index-hints.js
@@ -0,0 +1,383 @@
+/*jshint globalstrict:false, strict:false, maxlen: 500 */
+/*global fail, assertEqual, AQL_EXPLAIN */
+
+////////////////////////////////////////////////////////////////////////////////
+/// @brief tests for Ahuacatl, skiplist index queries
+///
+/// @file
+///
+/// DISCLAIMER
+///
+/// Copyright 2010-2016 ArangoDB GmbH, Cologne, Germany
+///
+/// Licensed under the Apache License, Version 2.0 (the "License");
+/// you may not use this file except in compliance with the License.
+/// You may obtain a copy of the License at
+///
+///     http://www.apache.org/licenses/LICENSE-2.0
+///
+/// Unless required by applicable law or agreed to in writing, software
+/// distributed under the License is distributed on an "AS IS" BASIS,
+/// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+/// See the License for the specific language governing permissions and
+/// limitations under the License.
+///
+/// Copyright holder is ArangoDB GmbH, Cologne, Germany
+///
+/// @author Michael Hackstein
+/// @author Copyright 2016, ArangoDB GmbH, Cologne, Germany
+////////////////////////////////////////////////////////////////////////////////
+
+var internal = require("internal");
+var db = internal.db;
+var jsunity = require("jsunity");
+var errors = internal.errors;
+
+////////////////////////////////////////////////////////////////////////////////
+/// @brief test suite
+////////////////////////////////////////////////////////////////////////////////
+
+function ahuacatlSkiplistOverlappingTestSuite () {
+  const getIndexNames = function (query) {
+    return AQL_EXPLAIN(query, {}, { optimizer: { rules: [ "-all", "+use-indexes" ] } })
+              .plan.nodes.filter(node => (node.type === 'IndexNode'))
+              .map(node => node.indexes.map(index => index.name));
+  };
+
+  const cn = 'UnitTestsIndexHints';
+  let collection;
+  let defaultEqualityIndex;
+  let alternateEqualityIndex;
+  let defaultSortingIndex;
+  let alternateSortingIndex;
+
+  return {
+
+////////////////////////////////////////////////////////////////////////////////
+/// @brief set up
+////////////////////////////////////////////////////////////////////////////////
+
+    setUp : function () {
+      internal.db._drop(cn);
+      collection = internal.db._create(cn);
+      
+      collection.ensureIndex({type: 'hash', name: 'hash_a', fields: ['a']});
+      collection.ensureIndex({type: 'hash', name: 'hash_a_b', fields: ['a', 'b']});
+      collection.ensureIndex({type: 'hash', name: 'hash_b_a', fields: ['b', 'a']});
+      collection.ensureIndex({type: 'skiplist', name: 'skip_a', fields: ['a']});
+      collection.ensureIndex({type: 'skiplist', name: 'skip_a_b', fields: ['a', 'b']});
+      collection.ensureIndex({type: 'skiplist', name: 'skip_b_a', fields: ['b', 'a']});
+      
+      const isMMFiles = db._engine().name === "mmfiles";
+      defaultEqualityIndex = isMMFiles ? 'skip_a' : 'hash_a';
+      alternateEqualityIndex = isMMFiles ? 'hash_a' : 'skip_a';
+      defaultSortingIndex = isMMFiles ? 'skip_a' : 'hash_a';
+      alternateSortingIndex = 'skip_a_b';
+    },
+
+    tearDown : function () {
+      internal.db._drop(cn);
+    },
+
+    testFilterNoHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {}
+          FILTER doc.a == 1
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], defaultEqualityIndex);
+    },
+
+    testFilterDefaultHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: '${defaultEqualityIndex}'}
+          FILTER doc.a == 1
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], defaultEqualityIndex);
+    },
+
+    testFilterDefaultHintForced : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: '${defaultEqualityIndex}', forceIndexHint: true}
+          FILTER doc.a == 1
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], defaultEqualityIndex);
+    },
+
+    testFilterNonexistentIndexHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: 'foo'}
+          FILTER doc.a == 1
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], defaultEqualityIndex);
+    },
+
+    testFilterNonexistentIndexHintForced : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: 'foo', forceIndexHint: true}
+          FILTER doc.a == 1
+          RETURN doc
+      `;
+      try {
+        const usedIndexes = getIndexNames(query);
+        fail();
+      } catch (err) {
+        assertEqual(errors.ERROR_QUERY_FORCED_INDEX_HINT_UNUSABLE.code, err.errorNum);
+      }
+    },
+
+    testFilterUnusableHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: 'hash_b_a'}
+          FILTER doc.a == 1
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], defaultEqualityIndex);
+    },
+
+    testFilterUnusableHintForced : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: 'hash_b_a', forceIndexHint: true}
+          FILTER doc.a == 1
+          RETURN doc
+      `;
+      try {
+        const usedIndexes = getIndexNames(query);
+        fail();
+      } catch (err) {
+        assertEqual(errors.ERROR_QUERY_FORCED_INDEX_HINT_UNUSABLE.code, err.errorNum);
+      }
+    },
+    
+    testFilterTypeHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: '${alternateEqualityIndex}'}
+          FILTER doc.a == 1
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], alternateEqualityIndex);
+    },
+
+    testFilterPartialCoverageHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: 'skip_a_b'}
+          FILTER doc.a == 1
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], 'skip_a_b');
+    },
+
+    testFilterListFirstHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: ['skip_a_b', '${alternateEqualityIndex}']}
+          FILTER doc.a == 1
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], `skip_a_b`);
+    },
+
+    testFilterListLastHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: ['skip_b_a', '${alternateEqualityIndex}']}
+          FILTER doc.a == 1
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], alternateEqualityIndex);
+    },
+
+    testFilterNestedMatchedHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: '${alternateEqualityIndex}'}
+          FILTER doc.a == 1
+          FOR sub IN ${cn} OPTIONS {indexHint: '${alternateEqualityIndex}'}
+            FILTER sub.a == 2
+            RETURN [doc, sub]
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 2);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], alternateEqualityIndex);
+      assertEqual(usedIndexes[1].length, 1);
+      assertEqual(usedIndexes[1][0], alternateEqualityIndex);
+    },
+
+    testFilterNestedUnmatchedHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: '${alternateEqualityIndex}'}
+          FILTER doc.a == 1
+          FOR sub IN ${cn} OPTIONS {indexHint: 'skip_a_b'}
+            FILTER sub.a == 2
+            RETURN [doc, sub]
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 2);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], alternateEqualityIndex);
+      assertEqual(usedIndexes[1].length, 1);
+      assertEqual(usedIndexes[1][0], 'skip_a_b');
+    },
+
+    testSortNoHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {}
+          SORT doc.a
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], defaultSortingIndex);
+    },
+
+    testSortDefaultHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: '${defaultSortingIndex}'}
+          SORT doc.a
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], defaultSortingIndex);
+    },
+
+    testSortDefaultHintForced : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: '${defaultSortingIndex}', forceIndexHint: true}
+          SORT doc.a
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], defaultSortingIndex);
+    },
+
+    testSortNonexistentIndexHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: 'foo'}
+          SORT doc.a
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], defaultSortingIndex);
+    },
+
+    testSortNonexistentIndexHintForced : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: 'foo', forceIndexHint: true}
+          SORT doc.a
+          RETURN doc
+      `;
+      try {
+        const usedIndexes = getIndexNames(query);
+        fail();
+      } catch (err) {
+        assertEqual(errors.ERROR_QUERY_FORCED_INDEX_HINT_UNUSABLE.code, err.errorNum);
+      }
+    },
+
+    testSortUnusableHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: 'skip_b_a'}
+          SORT doc.a
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], defaultSortingIndex);
+    },
+
+    testSortUnusableHintForced : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: 'skip_b_a', forceIndexHint: true}
+          SORT doc.a
+          RETURN doc
+      `;
+      try {
+        const usedIndexes = getIndexNames(query);
+        fail();
+      } catch (err) {
+        assertEqual(errors.ERROR_QUERY_FORCED_INDEX_HINT_UNUSABLE.code, err.errorNum);
+      }
+    },
+    
+    testSortPartialCoverageHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: '${alternateSortingIndex}'}
+          SORT doc.a
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], alternateSortingIndex);
+    },
+
+    testSortListFirstHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: ['${alternateSortingIndex}', '${defaultSortingIndex}']}
+          SORT doc.a
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], alternateSortingIndex);
+    },
+
+    testSortListLastHint : function () {
+      const query = `
+        FOR doc IN ${cn} OPTIONS {indexHint: ['skip_b_a', '${alternateSortingIndex}']}
+          SORT doc.a
+          RETURN doc
+      `;
+      const usedIndexes = getIndexNames(query);
+      assertEqual(usedIndexes.length, 1);
+      assertEqual(usedIndexes[0].length, 1);
+      assertEqual(usedIndexes[0][0], alternateSortingIndex);
+    },
+
+  };
+}
+
+////////////////////////////////////////////////////////////////////////////////
+/// @brief executes the test suite
+////////////////////////////////////////////////////////////////////////////////
+
+jsunity.run(ahuacatlSkiplistOverlappingTestSuite);
+
+return jsunity.done();