FactLane documentation
FactLane gives compatible AI-agent hosts a local memory layer for bounded facts: useful preferences, project facts, workflow rules, and other durable information that should survive longer than one conversation.
The important part is not merely remembering. FactLane separates a fact that is worth keeping from one that is trusted as Current, and it keeps memory subordinate to current instructions and live authoritative state.
New to FactLane?
Start with the outcome rather than the protocol:
- Read Everyday use cases to see what FactLane can remember and where its boundaries matter.
- Read the FAQ to check fit before spending time on setup.
- Read Core concepts for the plain-language model of Candidate, Current, scope, freshness, and authority.
- Follow the Quick start when you are ready to install and connect a supported local MCP host.
- Use Five MCP tools as the developer reference once the connection works.
Do I need to be a developer to benefit from it?
Not necessarily for normal use, but the current v0.1.3 setup is still technical.
FactLane is not a separate chatbot. It runs locally as a memory service behind a compatible AI agent host. Someone must install and configure that connection. After it is connected, the useful part can feel much simpler: keep using the assistant while approved preferences, project facts and recurring rules remain available across sessions.
If you are mainly trying to understand whether this solves your problem, start with Everyday use cases and the FAQ. You can leave Python, SQLite, MCP and embedding details until you are ready to install it yourself.
Building or integrating?
- Quick start — establish a safe read-only connection first.
- Five MCP tools — what each public tool is for after connectivity works.
- Architecture — request path, scope/identity model, storage and recovery.
- Environment and compatibility — Python, SQLite, embedding profiles and deployment limits.
- Release operations — exact release identity, upgrades, rollback and reverification.
- Security — trust boundaries and sensitive-memory recovery.
Current public release
FactLane v0.1.3 is the first official production release. Its qualified public profile is intentionally bounded: Python 3.11+, linked SQLite 3.42.0+, command-launched stdio MCP, supported local Ollama embeddings, and the documented local POSIX storage/recovery contract.
The live MCP schema and factlane --help-tools are authoritative for exact request signatures
and enum values.