> ## Documentation Index
> Fetch the complete documentation index at: https://retrievalcontextprotocol.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Memory

> Build a compact global or session memory over a corpus, then recall clues and entry-points that steer retrieval — MemoRAG and HippoRAG.

`memory/build` and `memory/recall` are gated by **`memory`**. Memory-augmented
RAG (MemoRAG, HippoRAG, long-term agent memory) builds a compact **global
memory** over a corpus, then at query time generates *clues* — surrogate
sub-queries or graph entry-points — that steer the actual retrieval. That
two-phase shape fits neither `retrieve` nor `index/add`, so `memory` makes it
explicit while staying fully optional.

## memory/build

Construct or update a memory over documents already indexed (or supplied
inline).

<CodeGroup>
  ```json Request theme={null}
  {
    "jsonrpc": "2.0", "id": 1, "method": "memory/build",
    "params": { "scope": "global" }
  }
  ```

  ```json Response theme={null}
  { "jsonrpc": "2.0", "id": 1, "result": { "memoryId": "mem-7", "tokens": 18240 } }
  ```
</CodeGroup>

<ParamField path="documents" type="array">
  `[{ id?, text }]` to build over; omit to build over the live index.
</ParamField>

<ParamField path="memoryId" type="string">
  Update an existing memory; omit to create a new one.
</ParamField>

<ParamField path="scope" type="&#x22;global&#x22; | &#x22;session&#x22;">
  Corpus-wide vs. trajectory-scoped memory.
</ParamField>

## memory/recall

Given a query and a memory, return **clues** the client then feeds into
`retrieve` / `graph`, plus optional draft `hits`.

<CodeGroup>
  ```json Request theme={null}
  {
    "jsonrpc": "2.0", "id": 2, "method": "memory/recall",
    "params": { "query": "how did the merger affect R&D spend?", "memoryId": "mem-7", "n": 5 }
  }
  ```

  ```json Response theme={null}
  { "jsonrpc": "2.0", "id": 2, "result": {
      "clues": [
        { "query": "R&D expenditure 2021 vs 2023", "weight": 0.9 },
        { "seedIds": ["e:acme", "e:merger"], "weight": 0.7 }
      ] } }
  ```
</CodeGroup>

<ParamField path="query" type="string" required>
  The user query to recall entry-points for.
</ParamField>

<ParamField path="memoryId" type="string">
  Which memory to recall against.
</ParamField>

<ParamField path="sessionId" type="string">
  Ties recall to an [agentic trajectory](/concepts/agentic-rag).
</ParamField>

<ParamField path="n" type="integer">
  Maximum number of clues to return.
</ParamField>

### Clue object

Each clue carries a surrogate `query`, a set of `seedIds` (graph entry nodes —
HippoRAG's personalized-PageRank seeds), or both, plus an optional `weight`. A
client fans clues out over [`retrieve`](/methods/retrieve) (usually with a
`sessionId`) or seeds a [`graph`](/methods/graph) traversal.

<Info>
  Memory is a heavyweight optional capability advertised as
  `memory: { scopes: [...], clues: true }`. Simple servers omit it entirely and
  lose nothing else — the rest of the protocol works unchanged.
</Info>