UNITARES
Self-hosted shared record of AI agent claims, reviews and outcomes for operators
- Pricing
- Open source
- Platforms
- self-hosted, macOS, API
- Version
- 3.6.0
- Listing updated
- today
Similar MCP servers
About UNITARES
UNITARES is an open-source MCP server for agent tooling. It is available on self-hosted, macOS, API and Linux.
The README frames UNITARES as a server you run that keeps a shared record of what autonomous agents claimed, what evidence they cited, who disputed it, and what replaced it. Agents connect over MCP or HTTP, so a Claude Code session, a Codex session and a later Hermes Agent run can all read and write the same log instead of relying only on Git history and handoff notes.
Each write is bound to the process that made it, and a newer finding can be linked as the replacement for an older one. Reviews and open objections stay attached to the work they concern, and test or build results are recorded against the check-ins that predicted them. A new session can search prior findings before acting on code it did not write.
The project is aimed at operators who run several long-lived agents across vendors on infrastructure they control. Installation is through a Homebrew preview for Apple Silicon Macs or a Docker Compose stack, and the core record is described as running without any external model API key.
What it does
- Shared agent record
- Findings, claims and outcomes from separate agent runs are stored in one operator-owned record.
- Supersession links
- When a finding is replaced, the older entry points to its successor so stale claims are visible.
- Attributed writes
- Every write is tied to the process that made it, and successors can declare which process they continue.
- Structured review trail
- Objections, conditions and their resolutions stay attached to the work they concern.
- Outcome recording
- Test and build results are logged against the check-ins that predicted them.
- Self-hosted deployment
- Runs on the operator's own machine or server through Homebrew or Docker Compose.
Good for
- Keep a record of which agent produced which finding across several coding sessions
- Let a later agent search prior reviews and open conditions before editing shared code
- Audit disputed agent changes and see what replaced them
- Coordinate multiple agents working in the same checkout with per-file edit leases
Pricing
Open source
Open source under Apache 2.0, self-hosted.
Prices change. Check cirwel.org before you buy.
UNITARES FAQ
Is UNITARES open source?
The website and README state it is released under the Apache License 2.0.
Does it need an external AI model API?
The README says the core record runs model-free with no API keys; an optional local model can be connected for consult and review.
Which agents can connect to it?
The README lists ready integrations for Claude Code, Codex and Hermes Agent, and says any agent that speaks MCP or HTTP can connect.
Is UNITARES free?
Yes. UNITARES is open source, so you can use and self-host it for free.
What are the best alternatives to UNITARES?
The closest alternatives to UNITARES are Agent Memory MCP and OpenCrawl.
Where this comes from
- Description: written from the project's README, the project's website and its MCP Registry entry.