UnisonIntroduction
Unison is the team brain for AI agents: a persistent, shared knowledge graph behind one HTTP API. Pick your surface and add memory in minutes — SDKs, a CLI, an MCP server, a filesystem mount, and framework integrations.
Unison gives a team of humans and AI agents one shared, persistent brain. Agents write decisions, facts, and documents into it; other agents — in other sessions, other harnesses, other machines — recall that context the moment they need it.
Mem0 remembers the user. Unison remembers the project.
Get started in your stack
The brain is one HTTP API. Reach it through a typed SDK, the CLI, an MCP server, a filesystem mount, or raw HTTP from any language.
TypeScript SDK
@unisonlabs/sdk — the typed brain client
Python SDK
unisonlabs on PyPI
CLI
the brain from your terminal
MCP server
native tools in Claude Code, Cursor, Codex
cURL / HTTP
any language, raw REST
unison-fs
mount the brain as a folder
Use it inside your agent framework
New to memory for agents?
Build a coding agent with memory
the recall → act → verify → write-back loop, with what the research shows
Add your first memory
persist and recall in three calls
How Unison works
the architecture in one page
What is a team brain?
and why it isn't per-user memory
Agent memory vs RAG
different questions — most teams need both
Why not just a vector store?
Bolting "memory" onto an agent — dump messages into a vector DB, embed, retrieve — gets you ~80%, and the model providers now give that away for free. The remaining 20% is where memory actually breaks, and it's what Unison is built for:
- Temporal correctness. Facts change. Unison stores bitemporal facts with supersession — it knows what changed when, and stops surfacing the version that's no longer true instead of returning a stale fact with confidence.
- Identity resolution. "Dan", "Daniel", and "the new backend hire" are one person. Unison runs tiered entity resolution so facts attach to the right entity instead of scattering across four.
- One brain, every agent. Per-tool memory is a silo. Unison is one source of truth every agent and teammate reads from and writes back to — so your fleet doesn't hold contradictory beliefs.
This is memory, not RAG — and we measure it in the open: unison-evals scores every memory system, including ours, on the same harness, with a decision-driving test most vendor benchmarks avoid.