Bilinc is a verifiable memory and state plane for autonomous agents.
It combines typed memory, belief revision, verification, audit trail, and recovery primitives so long-running agents can preserve durable state instead of only storing vectors.
Memory model
Working, episodic, procedural, semantic, and spatial memory.
Integrity
Cloud SDK, CLI, MCP setup, and hosted workflow documentation.
Recovery
Snapshots, diffs, and rollback for bounded incident response.
Four moves cover most agent-memory work.
Commit
Write a memory with type and importance through the hosted Cloud API.
Recall
Query memory through hosted recall profiles.
Revise
Deliberately correct a memory you already have, so the change is not an accidental overwrite.
Forget
Remove obsolete state from active recall with a required, audited reason.
Snapshot
Checkpoint a project before risky agent work.
Diff
Inspect what changed against a checkpoint, with values redacted by default.
Rollback
Restore a known-good checkpoint through preview plus explicit confirmation.
Status
Read the authenticated workspace, plan, limits, and capabilities of one API key.
Choose the retrieval posture, not just the query.
Bilinc fuses lexical, vector, and graph signals inside the hosted state plane. Current `main` also includes evidence-aware recall helpers such as claim projection, contradiction probes, and named recall profiles.
fast
Latency-first hosted recall.
balanced
Default profile for most agent workflows.
verified
Adds evidence metadata for operator review.
deep
Uses the richest hosted retrieval path when quality matters most.