Keel
Per-customer bitemporal context graph from product events, read by agents in five tools
- Pricing
- Open source
- Platforms
- CLI, API, self-hosted
- Runs
- Locally
- Version
- 0.38.0
- Listing updated
- 3 days ago
About Keel
Keel is an open-source MCP server for knowledge and memory. It runs locally on your machine and is started by your AI client.
Keel is meant for products that ship several AI agents and are tired of each one having its own data pipeline. You describe your domain once in a YAML cartridge of entities, events and rules; events arrive from Postgres CDC, an outbox, Kafka, SQS, webhooks or an SDK; and Keel builds one graph per tenant with no LLM in ingest, so replaying the same log yields the same graph.
Every fact records when it was true and when Keel learned it, so late events close facts rather than overwrite them, and reads accept as_of and as_known_at. Rules such as 'blocked' are evaluated in the graph, so the agent receives a verdict. The authors report 26 points higher accuracy on whole-project rule questions with this approach in their own benchmark, using fewer tokens.
The PyPI package keel-context runs over stdio and stores to memory, SQLite or MySQL via KEEL_DB; Postgres and DuckDB are also supported. Version 0.38.0 is labelled alpha.
Install
Run in a terminal
claude mcp add keel -- uvx keel-contextAdd to claude_desktop_config.json
{
"mcpServers": {
"keel": {
"command": "uvx",
"args": [
"keel-context"
]
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"keel": {
"command": "uvx",
"args": [
"keel-context"
]
}
}
}Add to .vscode/mcp.json
{
"servers": {
"keel": {
"type": "stdio",
"command": "uvx",
"args": [
"keel-context"
]
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"keel": {
"command": "uvx",
"args": [
"keel-context"
]
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"keel": {
"command": "uvx",
"args": [
"keel-context"
]
}
}
}Configuration
KEEL_DBA SQLite file or a mysql:// URL; in memory when unsetKEEL_AGENTThe agent id, when the cartridge declares agents
Tools this server provides (5)
Taken from the project documentation; the running server may differ.
- context
- Returns what is known about a subject now or at a past moment, with facts, relations and rule verdicts.
- history
- Shows how a subject changed over time and when it was learned.
- search
- Finds subjects by text, kind, rule, attribute value or meaning.
- related
- Returns connected subjects up to three hops away.
- annotate
- Records a conclusion by an agent or person with evidence and scope.
What it does
- Cartridges
- Declare entities, events, rules and memory policy in YAML.
- Bitemporal facts
- Tracks valid time and recorded time for every fact.
- Rules evaluated in-graph
- Agents receive verdicts like 'blocked' instead of deriving them.
- Deterministic ingest
- No LLM in ingest; replay gives the same graph.
- Provenance
- Source, evidence and corrections travel with each fact.
- Multiple stores
- In-memory, SQLite, MySQL, Postgres or DuckDB.
Good for
- Give several in-product agents one shared view of a customer
- Ask what was true for a tenant at a past date
- Audit what another agent knew when it decided
- Ingest Postgres changes as events
- Record agent conclusions with evidence and scope
Pricing
Open source
Apache 2.0; hosted console in beta
Prices change. Check angelkurten.github.io before you buy.
Keel FAQ
What are the five tools?
context, history, search, related and annotate; an optional read-only SQL `query` tool is off by default.
What licence?
Apache 2.0.
What is its maturity?
Version 0.38.0, labelled alpha.
Can I use it without MCP?
Yes, through an HTTP JSON API and SDKs for TypeScript, Python, Go and PHP.
Is Keel free?
Yes. Keel is open source, so you can use and self-host it for free.
How do I add Keel to Claude Code?
Run this command in your terminal: claude mcp add keel -- uvx keel-context
What are the best alternatives to Keel?
The closest alternatives to Keel are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.
Alternatives to Keel

REA
MCP serverCLI and MCP server that lets coding agents reverse engineer apps and binaries
Runs locally- Open source
- 464K GitHub stars
- Developer tools

Codebase Memory
MCP serverLocal MCP server that builds a code knowledge graph for AI coding agents
Runs locally- Open source
- 48.7K GitHub stars
- Knowledge and memory

GitHub
MCP serverOfficial MCP server connecting AI assistants to GitHub repos, issues and workflows
Not checked yet- Remote
- Open source
- 34.5K GitHub stars
- Developer tools

Serena MCP: the IDE for your agent
MCP serverMCP toolkit giving coding agents symbol-level code retrieval and editing
Runs locally- Open source
- 31.5K GitHub stars
- Developer tools

XERJ
MCP serverLocal search engine for AI agents with cited answers over your code and docs
Runs locally- Open source
- 19.2K GitHub stars
- Knowledge and memory

Ouroboros
MCP serverSpec-first layer that pins requirements and verifies results of AI coding agent work
Runs locally- Open source
- 6.4K GitHub stars
- Developer tools