Skip to main content
Glama
MCAP MCP Server

MCAP MCP Server

by turkenberg

MCAP MCP Server

Query your robot's MCAP recordings with SQL — straight from your LLM.

codecov License

Setup

{
  "mcpServers": {
    "mcap-query": {
      "command": "uvx",
      "args": ["mcap-mcp-server[all]"]
    }
  }
}

No install, no database, no API keys. Requires uv.

Related MCP server: MCP Data Analyst

Usage

Just talk to your LLM:

  • "Tell me what topics are in session_003.mcap"

  • "In session_017.mcap find all moments where voltage dropped below 22V"

  • "Correlate IMU acceleration with motor current."

  • "Compare average battery voltage across my last 5 runs"

  • "What version of mcap-mcp-server am I running? Update it"

Tools

Tool

Needs loading

What it does

list_recordings

no

Find MCAP files in your project (or any path)

get_recording_info

no

Metadata, channels, attachments for a file

get_schema

no

SQL table names & column types — for query planning

load_recording

—

Decode MCAP into DuckDB (the LLM calls this automatically)

query

yes

Run SQL (full DuckDB — including ASOF JOIN)

get_version

no

Server version, available decoders, upgrade command

Project documentation — configuration, Docker, development setup, and architecture.

Example SQL (under the hood)

-- Time-windowed stats
SELECT (timestamp_us / 1000000) as second,
       AVG(voltage) as avg_v, MIN(voltage) as min_v
FROM battery GROUP BY second ORDER BY second

-- Cross-sensor correlation via ASOF JOIN
SELECT b.timestamp_us, b.voltage, i.linear_acceleration_x
FROM battery b ASOF JOIN imu i ON b.timestamp_us >= i.timestamp_us

-- Multi-recording comparison
SELECT 'run1' as run, AVG(voltage) FROM r1_battery
UNION ALL
SELECT 'run2', AVG(voltage) FROM r2_battery

Performance

Metadata tools (list_recordings, get_recording_info, get_schema) return in < 1 ms regardless of file size. SQL queries execute in 1–20 ms once data is loaded. The one-time load_recording cost scales with file size:

Messages

File size

Load time

Memory

Query time

1K

23 KB

8 ms

< 1 MB

1 ms

10K

220 KB

90 ms

0.5 MB

1–3 ms

100K

2.2 MB

0.7 s

5 MB

1–5 ms

500K

11 MB

3.9 s

23 MB

2–9 ms

1M

23 MB

8 s

46 MB

2–13 ms

2M

48 MB

18 s

92 MB

2–22 ms

Measured on Apple M4 with JSON-encoded messages, 5 fields per message. Query times are median across aggregation, filter, and window function queries. Memory is the DuckDB in-memory footprint (default budget: 2 GB).

Tip: use topics and start_time/end_time filters on load_recording to load only what you need.

Update

uvx mcap-mcp-server[all] --upgrade

Or ask your LLM — the get_version tool returns the running version and the upgrade command.

License

GNU General Public License v3.0 — see LICENSE.

Available Tools

6 tools
get_recording_infoA

Get full metadata, channel details, and attachment list for a specific MCAP recording file. Does not require loading. Use this for detailed inspection before loading data.

Inputs: file (required). Returns: result

get_schemaA

Inspect the SQL schema for a recording: topic names, table names, column names and DuckDB types. Does not require loading. Use this to plan SQL queries before running them. Returns a sql_hint with JOIN guidance.

Inputs: file (required), topic. Returns: result

get_versionA

Return the server version, supported encodings, and upgrade command. Use this to check for updates or diagnose compatibility issues.

Inputs: no named parameters. Returns: result

list_recordingsA

Discover available MCAP recording files. Does not require loading. Returns file names, sizes, durations, channel lists, and message counts. Use this first to see what data is available before loading. By default scans the project directory; pass an absolute 'path' to scan any directory on the filesystem.

Inputs: path, after, before. Returns: result

load_recordingA

Decode an MCAP file and load its data into DuckDB for SQL querying. This decodes all messages and may take seconds to tens of seconds depending on file size. You must call this before running queries. For large files, use 'topics' to load only the topics you need and 'start_time'/'end_time' to narrow the time window — this significantly reduces both load time and memory usage. Set an alias for multi-recording comparison.

Inputs: file (required), alias, topics, end_time, +2 more. Returns: result

queryA

Execute a SQL query against loaded MCAP data. Supports full DuckDB SQL including JOINs, GROUP BY, window functions, and ASOF JOIN for time-series correlation. Data must be loaded first via load_recording. If a table is missing, call load_recording with the needed topic.

Inputs: sql (required), limit. Returns: result

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4.4/5.0

    Scored across 6 tools

    Disambiguation5/5

    Each tool has a clearly distinct purpose: listing, inspecting, schema inspection, version info, loading, and querying. There is no ambiguity or overlap.

    Naming Consistency5/5

    All tool names follow a consistent verb_noun snake_case pattern (e.g., list_recordings, load_recording). Even 'query' fits as a single verb without a noun, maintaining uniformity.

    Tool Count5/5

    With 6 tools, the server is well-scoped for its domain of MCAP recording management. Each tool serves a necessary step in the workflow without redundancy.

    Completeness5/5

    The tool surface covers the full lifecycle: discovery, inspection, loading, and querying. No obvious gaps exist for typical usage scenarios.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      A
      maintenance
      Enables AI assistants to query databases using natural language, with automatic schema discovery and SQL compilation.
      613 npm
      3,238
      Apache 2.0
    • A
      license
      Not graded
      quality
      D
      maintenance
      Enables natural language querying of SQL databases using AI, supporting multiple database types and automatic schema discovery.
      1
      MIT
    • F
      license
      Not graded
      quality
      D
      maintenance
      Enables natural language querying of SQL databases with robust safety guarantees including read-only enforcement, AST validation, and row caps.
      -