MCP / Claude Integration¶

agentdelta ships an MCP (Model Context Protocol) server that exposes its core operations as native Claude tools.

Install¶

pip install "agentdelta[mcp]"

Edit ~/.config/claude/claude_desktop_config.json (Linux) or
~/Library/Application Support/Claude/claude_desktop_config.json (macOS):

{
  "mcpServers": {
    "agentdelta": {
      "command": "agentdelta-mcp"
    }
  }
}

Restart Claude Desktop. You now have three tools available in every conversation:

Compare two JSONL trace files and return a DiffResult JSON.

diff_traces(trace_a="/path/to/baseline.jsonl", trace_b="/path/to/candidate.jsonl")

Returns: has_regression, fork_point (step, description, similarity), summary stats, and per-step alignment.

Summarise a single trace file's execution path.

inspect_trace(trace_path="/path/to/trace.jsonl")

Returns: run_id, node/edge count, step breakdown with type and content preview.

Get a copy-paste Python snippet to record an agent run.

record_snippet(framework="langchain")

Returns: ready-to-run Python code for the specified framework.

After cloning the repo, these project-level commands are available in Claude Code:

Command	What it does
`/project:diff <baseline> <candidate>`	Diff two traces and explain the fork
`/project:inspect <trace>`	Summarise a trace's execution path
`/project:record [framework]`	Generate recording boilerplate
`/project:add-adapter <framework>`	Scaffold a new framework adapter
`/project:pr-prep`	Run lint + types + tests + CHANGELOG check
`/project:test`	Run test suite and report failures

agentdelta is also listed on smithery.ai — an MCP server marketplace. Search for "agentdelta" to install with one click from Claude Desktop.