Agent frameworks

Memory for Agent Frameworks

Give your LangChain, CrewAI, Mastra, or AutoGPT agent persistent graph memory via REST.

Supported frameworks

LangChain / LangGraph

Wrap POST /ingest and POST /search as tools. Register with your agent executor or LangGraph node.

CrewAI

Expose Graphory as Tools on any crew member. Shared graph memory across agents in a crew.

Mastra

TypeScript SDK via REST. Call the Graph API from any Mastra workflow or tool step.

AutoGPT

Memory plugin pattern. Replace the default memory store with Graphory's persistent graph.

Pydantic AI

Define Graphory search and write tools as Pydantic AI tool functions. Type-safe memory, free.

Custom agents

Any REST client works. POST /ingest to remember, POST /search to recall, POST /traverse to explore.

One REST call to remember.

Same API for every framework. Save with POST /ingest, recall with POST /search.

Remember

curl -X POST https://api.graphory.io/ingest \
  -H "Authorization: Bearer gs_ak_your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "type": "note",
    "title": "Acme pricing blocker",
    "body": "Sarah at Acme flagged pricing as the blocker in March.",
    "date_published": "2026-03-14"
  }'

Recall

curl -X POST https://api.graphory.io/search \
  -H "Authorization: Bearer gs_ak_your_key" \
  -H "Content-Type: application/json" \
  -d '{"query": "Acme pricing"}'

Full reference at docs.graphory.io. Or use the Python SDK: pip install graphory.

Why graph memory for agents

Shared memory across agents

Multi-agent crews write to one graph. Every agent reads the same truth - no message-passing gymnastics.

Traversable

Your agent can follow real business relationships. Not fuzzy retrieval, actual paths.

Deterministic

No embeddings, no hallucinated recall. Exact queries, exact results. Reproducible runs.

Temporal

Every memory carries a timestamp, a source, and an audit trail. Audit every decision, replay every run.

Measured, not marketed.

Deterministic extraction and retrieval, benchmarked against public datasets.

0.9107 accuracy

LongMemEval (agent + MCP)

0.8667 accuracy

LoCoMo-MC10 (agent + MCP)

0.900 F1

BizLineItemBench

One REST API. Every agent framework.

Free up to 100K nodes. No credit card.