# CommitLayer

> CommitLayer is a freemium MCP server for agent tooling. MCP server that lets AI assistants read verdicts and evidence for recorded agent actions. It runs as a hosted remote server, so there is nothing to install locally.

- **Kind:** MCP server
- **Pricing:** Freemium
- **Starting price:** $99 a month
- **Platforms:** API, Web
- **Website:** https://commitlayer.ai/mcp
- **MCP Registry name:** ai.commitlayer/mcp
- **Categories:** agent tooling
- **Listing updated:** 2026-11-18

## About

The server gives an assistant read access to CommitLayer's records of what agents did, the verdict on each action, the reasons behind it, and the evidence pointers that back it. The site describes the split clearly: the assistant reads verdicts and reasons, while the agent declares its intended actions and the verdict still comes from the customer's systems of record.

Beyond reading, the server exposes contract tools. An assistant can validate a contract, dry-run it over a month of claims, or draft a proposed contract from recorded actions, but nothing is saved through MCP; a person saves contracts in the app. Write access is limited to declaring an action before it happens, which needs the actions:record scope.

It suits teams running AI support agents who want an assistant to investigate failed actions or review contract changes without being able to alter facts, contracts or verdicts. Access is set by API key presets or OAuth with a workspace-bound token.

## Features

- **Action verdict lookup:** Retrieves a recorded action with its status, reason codes, checks run and evidence pointers.
- **Pending action tracking:** Lists actions whose verdict is still provisional and the sources they are waiting on.
- **Contract validation and dry runs:** Validates contracts and runs them over a month's claims without saving anything.
- **Connector health reporting:** Shows status, trust level and last sync for each connection, without exposing credentials.
- **Hash verification:** Checks whether a statement or verdict with a given hash exists and whether its chain is intact.
- **Audit-logged calls:** Every tool call is written to the workspace audit log with the principal that made it.

## Good for

- Investigating why an agent action failed by tracing from verdict to evidence
- Reviewing a proposed contract change with dry-run results before it is saved
- Tracking actions still awaiting settlement from external systems
- Declaring an agent's intended action before it runs, through a recorder key

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http mcp https://api.commitlayer.ai/mcp
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://api.commitlayer.ai/mcp
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "mcp": {
      "url": "https://api.commitlayer.ai/mcp"
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "mcp": {
      "type": "http",
      "url": "https://api.commitlayer.ai/mcp"
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "mcp": {
      "serverUrl": "https://api.commitlayer.ai/mcp"
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "mcp": {
      "type": "http",
      "url": "https://api.commitlayer.ai/mcp"
    }
  }
}
```

## Tools

- `get_action`: One recorded action with both lenses, status, reason codes, the checks that ran and evidence pointers.
- `list_actions`: Recorded actions by contract, workflow, verdict or month, each with its latest verdict.
- `pending_actions`: Actions whose verdict is still provisional, with the sources they wait on and when they settle.
- `workflow_summary`: Per-contract counts for a month: pass, fail, unknown, pending, compensated.
- `record_action`: Declare an action before it is performed and return the provisional policy verdict.
- `list_contracts`: The workspace's contracts and the library's.
- `get_contract`: A single contract with its YAML, compiled rules, plain language and versions.
- `validate_contract`: Schema errors, compiled rules, plain-language rendering and weakening warnings without saving.
- `dry_run_verify`: What a saved or proposed contract would decide over a month's claims and facts, without persisting.
- `draft_contract`: A proposed contract drafted from recorded actions and observed effects, without saving.
- `verify_hash`: Whether a statement, verdict or dispute package with a given hash exists and its chain is intact.
- `connector_health`: Status, trust level, provided data, last sync and watermark for every connection.
- `list_reconciliations`: A vendor's claims for a period, listed with coverage and totals.
- `get_reconciliation`: Coverage, claims, run and statement totals for a single reconciliation.
- `usage`: Plan, limits, usage this month and the allowance of verified actions.

## FAQ

### Can the assistant change facts, contracts or verdicts through this server?

No. The site says no scope grants writing or editing facts, saving or activating contracts, or overriding verdicts. A model-facing key or OAuth token can declare and read.

### How do I connect a client?

With an API key, send a Read-only key as a bearer token to https://api.commitlayer.ai/mcp. Remote clients such as Claude.ai can use OAuth, which asks you to approve one workspace and scopes.

### Do MCP calls count toward usage limits?

Yes. The site says an action recorded over MCP counts toward the monthly allowance like any other.

### Is CommitLayer free?

CommitLayer has a free plan, with paid plans for more usage or features. Paid plans start at $99 a month.

### How do I add CommitLayer to Claude Code?

Run this command in your terminal: claude mcp add --transport http mcp https://api.commitlayer.ai/mcp

### What are the best alternatives to CommitLayer?

The closest alternatives to CommitLayer are BotShelf Vampire, Agent interoperability design — Atom, Agent context packaging — Atom, Zero knowledge claim audit — Atom and Zero trust evaluation verdict — Atom.

## Where this comes from

- Description: written from the project's website and its MCP Registry entry.
- Install steps: from the official MCP Registry entry.

## Alternatives

- [BotShelf Vampire](https://ai-powered.com/mcp/botshelf-vampire): BotShelf Vampire is an open-source MCP server for agent tooling.
- [Agent interoperability design — Atom](https://ai-powered.com/mcp/agent-interoperability-design-atom): Agent interoperability design — Atom is an MCP server for design and media.
- [Agent context packaging — Atom](https://ai-powered.com/mcp/agent-context-packaging-atom): Agent context packaging — Atom is an MCP server for agent tooling.
- [Zero knowledge claim audit — Atom](https://ai-powered.com/mcp/zero-knowledge-claim-audit-atom): Zero knowledge claim audit — Atom is an MCP server for security.
- [Zero trust evaluation verdict — Atom](https://ai-powered.com/mcp/zero-trust-evaluation-verdict-atom): Zero trust evaluation verdict — Atom is an MCP server for ai models and llm tooling.
- [Network exposure audit — Atom](https://ai-powered.com/mcp/network-exposure-audit-atom): Network exposure audit — Atom is a paid MCP server for security.
- [BotMarket](https://ai-powered.com/mcp/botmarket): BotMarket is a free MCP server for agent tooling.
- [WritersLogic agent-infra](https://ai-powered.com/mcp/writerslogic-agent-infra): WritersLogic agent-infra is a paid MCP server for agent tooling.

---
Source: https://ai-powered.com/mcp/commitlayer
