agentleFS
all-source-osall-source-os/all-source/.claude/skills/allsource-mcp-setup/SKILL.md

allsource-mcp-setup

Skill4 starsChanged yesterdayApache-2.0View on GitHub

Configure the AllSource MCP server for local debugging in Claude Code. Triggers on: "configure allsource mcp", "setup mcp", "allsource mcp config", "add allsource to claude code", "mcp server setup". Reads directly from WAL + Parquet files. No running server needed. Binary at: target/release/allsource-mcp Add to ~/.claude/settings.json or project .claude/settings.json: After configuration, these tools appear in Claude Code: After adding the config, restart Claude Code and check: 1. Run /mcp — allsource should appear in the server list 2. Ask Claude…

The file

Show its rules and text

We split this skill into the rules it gives an agent

17 rules. About 0 are common, 9 are distinctive, and 8 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 8

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Add to ~/.claude/settings.json or project .claude/settings.json:

    line 24
  2. Distinctive:

    After configuration, these tools appear in Claude Code:

    line 67
  3. Distinctive:

    After adding the config, restart Claude Code and check:

    line 82
  4. Very specific:

    Run /mcp — allsource should appear in the server list

    line 83
  5. Distinctive:

    Ask Claude to "use quick_stats" — should return event counts

    line 84
  6. Distinctive:

    Ask Claude to "query events for entity X" — should return events

    line 85
  7. Very specific:

    If the Docker compose stack is up (allsource-core-leader on port 3280):

    line 114
  8. Very specific:

    "No such file or directory": Check the binary path. Build with cargo build --release -p allsource-mcp.

    line 160
  9. Distinctive:

    Permission denied: Check file permissions on the data directory.

    line 162
  10. Very specific:

    Verify the data directory has files: ls <data-dir>/storage/ should show .parquet files and/or ls <data-dir>/wal/ should show .log files.

    line 166
  11. Distinctive:

    If empty, the application may not have written events yet, or the path is wrong.

    line 167
  12. Very specific:

    Check if the data dir has a __system/ subdirectory — if yes, it's an AllSource data dir.

    line 168
  13. Very specific:

    Run /mcp in Claude Code to check server status.

    line 172
  14. Very specific:

    Check stderr output: set RUST_LOG=debug in the env to see MCP protocol messages.

    line 173
  15. Distinctive:

    Ensure the JSON in settings is valid (no trailing commas).

    line 174
  16. Distinctive:

    After first open, subsequent queries are fast (data cached in DashMap).

    line 180
  17. Distinctive:

    If the application is still running and writing, new events won't appear until MCP server restarts.

    line 185

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

AllSource MCP Setup

Configure the AllSource MCP server for local debugging in Claude Code.

Triggers on: "configure allsource mcp", "setup mcp", "allsource mcp config", "add allsource to claude code", "mcp server setup".


Option 1: Rust stdio MCP server (recommended for local debugging)

Reads directly from WAL + Parquet files. No running server needed.

Build

cd <chronos-monorepo>
cargo build --release -p allsource-mcp

Binary at: target/release/allsource-mcp

Claude Code Configuration

Add to ~/.claude/settings.json or project .claude/settings.json:

{
  "mcpServers": {
    "allsource": {
      "command": "<path-to>/allsource-mcp",
      "args": ["--data-dir", "<path-to-data>"]
    }
  }
}

Longhand on macOS

{
  "mcpServers": {
    "allsource": {
      "command": "/Users/<you>/Projects/chronos/chronos-monorepo/target/release/allsource-mcp",
      "args": ["--data-dir", "/Users/<you>/Library/Application Support/Longhand/allsource"]
    }
  }
}

Using environment variable

{
  "mcpServers": {
    "allsource": {
      "command": "/path/to/allsource-mcp",
      "env": {
        "ALLSOURCE_DATA_DIR": "/path/to/data"
      }
    }
  }
}

Available Tools

After configuration, these tools appear in Claude Code:

ToolDescription
query_eventsQuery events (entity_id, event_type, time range, limit)
sample_eventsSample recent events across all entities
quick_statsStore summary: counts, types, date range, durability
get_snapshotProjection/snapshot state for an entity
event_timelineChronological event list for an entity
explain_entityHuman-readable entity lifecycle summary
reconstruct_stateFold events to rebuild current state
analyze_changesChanges within a time window

Verification

After adding the config, restart Claude Code and check:

  1. Run /mcp — allsource should appear in the server list
  2. Ask Claude to "use quick_stats" — should return event counts
  3. Ask Claude to "query events for entity X" — should return events

Option 2: Elixir MCP server (remote mode)

Connects to a running Core server over HTTP. Use when Core is running in Docker.

Configuration

{
  "mcpServers": {
    "allsource": {
      "command": "mix",
      "args": ["phx.server"],
      "cwd": "<chronos-monorepo>/apps/mcp-server-elixir",
      "env": {
        "CORE_URL": "http://localhost:3900",
        "CORE_MODE": "remote",
        "PORT": "3904"
      }
    }
  }
}

With Docker stack running

If the Docker compose stack is up (allsource-core-leader on port 3280):

{
  "mcpServers": {
    "allsource": {
      "command": "mix",
      "args": ["phx.server"],
      "cwd": "<chronos-monorepo>/apps/mcp-server-elixir",
      "env": {
        "CORE_URL": "http://localhost:3280",
        "CORE_MODE": "remote"
      }
    }
  }
}

Option 3: Elixir MCP server (embedded mode)

Runs Core in-process via Rustler NIF. No separate Core needed, but requires Rust + Elixir toolchain.

{
  "mcpServers": {
    "allsource": {
      "command": "mix",
      "args": ["phx.server"],
      "cwd": "<chronos-monorepo>/apps/mcp-server-elixir",
      "env": {
        "CORE_MODE": "embedded",
        "CORE_DATA_DIR": "/path/to/data"
      }
    }
  }
}

Troubleshooting

Server not starting

  • "No such file or directory": Check the binary path. Build with cargo build --release -p allsource-mcp.
  • "data_dir is required": Pass --data-dir arg or set ALLSOURCE_DATA_DIR env var.
  • Permission denied: Check file permissions on the data directory.

No events found

  • Verify the data directory has files: ls <data-dir>/storage/ should show .parquet files and/or ls <data-dir>/wal/ should show .log files.
  • If empty, the application may not have written events yet, or the path is wrong.
  • Check if the data dir has a __system/ subdirectory — if yes, it's an AllSource data dir.

Tools not appearing

  • Run /mcp in Claude Code to check server status.
  • Check stderr output: set RUST_LOG=debug in the env to see MCP protocol messages.
  • Ensure the JSON in settings is valid (no trailing commas).

Slow startup

  • Large Parquet collections take time to scan on first open.
  • WAL replay happens at startup — many WAL entries = slower start.
  • After first open, subsequent queries are fast (data cached in DashMap).

Wrong data / stale results

  • The MCP server opens the data directory at startup and replays WAL.
  • If the application is still running and writing, new events won't appear until MCP server restarts.
  • For live debugging, use the Elixir MCP server in remote mode connected to the running Core.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.