
Attestor
Self-hosted memory layer that keeps agent context flat at around 200 tokens per call instead of re-reading the entire conversation history each turn. Exposes `add` and `recall` operations over stdio, backed by Postgres for document storage, Pinecone Local for vector retrieval, and Neo4j for entity graphs. Every memory carries bi-temporal timestamps so you can replay any past state, and the conflict resolver logs an auditable supersession chain when facts get updated or invalidated. Reach for this when you're running multi-agent teams that share memory across sessions or need point-in-time reconstruction for compliance. The quickstart command brings up the three-container local stack and wires the MCP server into Claude Desktop in one shot.
- GitHub stars
- 13
- tools
- 8
- auth
- Secret needed
- transport
- STDIO
- env vars
- 4
- last updated
- Apr 29, 2026
Tools
8 toolsPublic tool metadata for what this MCP can expose to an agent.
memory_addWrite a memory with provenance
Write a memory with provenance
No parameter schema in public metadata yet.
memory_getFetch one memory by id
Fetch one memory by id
No parameter schema in public metadata yet.
memory_recallRun the full retrieval pipeline
Run the full retrieval pipeline
No parameter schema in public metadata yet.
memory_searchFiltered list (entity / category / time / namespace)
Filtered list (entity / category / time / namespace)
No parameter schema in public metadata yet.
memory_forgetArchive a memory by id
Archive a memory by id
No parameter schema in public metadata yet.
memory_timelineChronology for an entity
Chronology for an entity
No parameter schema in public metadata yet.
memory_statsStore statistics
Store statistics
No parameter schema in public metadata yet.
memory_healthPer-role health snapshot — call this first when integrating
Per-role health snapshot — call this first when integrating
No parameter schema in public metadata yet.
Sponsors
Configuration
ATTESTOR_PATHDirectory containing attestor's config.toml. Defaults to ~/.attestor.
ATTESTOR_DISABLE_LOCAL_EMBEDSet to 1 to skip the local Ollama embedder probe (e.g., on hosts without Ollama). Defaults to unset (Ollama is the default embedder).
OPENAI_API_KEYsecretOpenAI API key (used for the OpenAI text-embedding-3-large fallback embedder when Ollama is unavailable).
OPENROUTER_API_KEYsecretOpenRouter API key (used for the federated LLM extraction / consolidation paths and for the LongMemEval benchmark runner).
Readme
Attestor
Governed memory for multi-agent production meshes.
Attestor is no longer "memory for your coding assistant." Three things happened: the editor vendors shipped memory for free, the benchmark race became a race about which LLM answers the question, and every serious conversation about a memory layer was about something else entirely: when many agents share one memory, who is allowed to write, who is allowed to forget, and can you prove it afterwards. So that is what Attestor is now.
Details
- Repository
- github.com/bolnet/attestor
- Categories
- AI & LLM Tools, Communication & Messaging, Automation & Workflows, Monitoring & Observability
- Package
attestor- Transport
- STDIO
- Auth
- Required
- Registry
- MCP registry, active