keelyard
Proof gate, edit snapshots and output trimming for Claude Code, Codex and OpenCode
- Pricing
- Open source
- Platforms
- CLI
- Runs
- Locally
- Version
- 1.2.1
- Listing updated
- 2 days ago
About keelyard
keelyard is an open-source MCP server for developer tools. It runs locally on your machine and is started by your AI client.
keelyard sits between a coding agent and the repository. Its proof gate stops a session from finishing until there is a checklist, evidence for each item and a passing test run after the last edit, and it writes a one-file HTML report. Every edit is snapshotted in a separate git repo, so keelyard rewind to <n> undoes a step and you can fork from any one without touching your own .git.
It also shortens tool output: filters keep every failure while removing repeated passing lines, reads are deduplicated, and project_map, outline and find_symbol tools reduce what the agent has to read. Further pieces cover memory search across past sessions, token and cost reports from local logs, a Swarm mode that offloads bulk reading to a cheaper model with citations checked, and an Arena for comparing agents on one task.
The project reports one measured result: on OpenCode with a free model, the lean preset cut session input tokens 14.2x on a noisy 1,566-test suite and 6.0x on a small one, and says Claude Code and Codex savings are not yet measured. It runs locally with no telemetry under Apache-2.0; keelyard init wires up hooks, MCP entries and guide files.
Install
Run in a terminal
claude mcp add keelyard -- uvx keelyardAdd to claude_desktop_config.json
{
"mcpServers": {
"keelyard": {
"command": "uvx",
"args": [
"keelyard"
]
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"keelyard": {
"command": "uvx",
"args": [
"keelyard"
]
}
}
}Add to .vscode/mcp.json
{
"servers": {
"keelyard": {
"type": "stdio",
"command": "uvx",
"args": [
"keelyard"
]
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"keelyard": {
"command": "uvx",
"args": [
"keelyard"
]
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"keelyard": {
"command": "uvx",
"args": [
"keelyard"
]
}
}
}Tools this server provides (3)
Taken from the project documentation; the running server may differ.
- project_map
- Overview of the project structure.
- outline
- Outline of a file.
- find_symbol
- Locate a symbol in the codebase.
What it does
- Proof gate
- Blocks the agent finishing until a passing test run follows its last edit.
- Rewind snapshots
- Snapshots every edit in a separate git repo for undo and forking.
- Output trimming
- Filters long logs but keeps every failure.
- Context navigation tools
- project_map, outline and find_symbol reduce file reading.
- Memory and usage
- Searches past sessions and reports tokens and cost from local logs.
- One-command setup
- `keelyard init` configures hooks, MCP and guides for detected agents.
Good for
- Stop an agent from saying done while tests fail
- Roll back a bad agent edit, including shell-made changes
- Cut token use on noisy test output
- Compare several agents on the same task
- Search what earlier sessions did
Pricing
Open source
Free, Apache-2.0
Prices change. Check akobirjontulqinovfuture.github.io before you buy.
keelyard FAQ
Which agents does it support?
Claude Code, Codex and OpenCode.
Does it send data anywhere?
No. It runs locally with no telemetry and is not a proxy; it only uses the network if you configure a helper model for swarm.
Is it open source?
Yes, Apache-2.0.
How much does it save?
The README reports measured OpenCode savings of 14.2x and 6.0x and says Claude Code and Codex savings are not measured yet.
Is keelyard free?
Yes. keelyard is open source, so you can use and self-host it for free.
How do I add keelyard to Claude Code?
Run this command in your terminal: claude mcp add keelyard -- uvx keelyard
What are the best alternatives to keelyard?
The closest alternatives to keelyard are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.
Alternatives to keelyard

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