Fw Context MCP Server

fw-context

mcp-name: io.github.turbyho/fw-context-mcp

MCP server for embedded C/C++ firmware — gives AI assistants (Claude Code, Cursor, OpenCode, etc.) real understanding of your codebase. Parses your actual build with libclang, extracts every symbol, and builds a persistent index with full-text search, call graph, and vector embeddings.

No hallucination. No grepping. No reading thousands of framework headers into context.

What it does

Your AI assistant goes from guessing to knowing:

"What does uart_init do and who calls it?" → get_symbol_context("uart_init") — body, callers, callees in one call.

"Find all BLE advertising functions and how they're connected." → search_code("ble advertising", kind="function") → find_call_path("gap_init", "start_advertising")

"Show me the implementation of adc_read — not the declaration." → get_source("adc_read") — exact body via libclang, no file reading.

"What would break if I change spi_transfer?" → find_all_callers_recursive("spi_transfer") — every caller, direct and indirect.

"Give me a map of modem_msg.cpp before I read it." → get_file_map("src/modem_msg.cpp") — 426 symbols grouped by kind.

19 MCP tools — symbol search, source reading, call-graph traversal, hotspot analysis, dead code detection, vector search. All backed by real compiler flags from compile_commands.json — #ifdef-aware, not grep.

Quick start

# 1. Install
git clone git@github.com:turbyho/fw-context-mcp.git ~/.fw-context/src
uv venv ~/.fw-context/.venv --python 3.12
uv pip install --python ~/.fw-context/.venv/bin/python ~/.fw-context/src/
echo 'export PATH="$HOME/.fw-context/.venv/bin:$PATH"' >> ~/.zshrc

# 2. Register with your AI assistant
fw-context init

# 3. Generate compile_commands.json and index
cd your-firmware-project

# Mbed OS:
bear -- mbed compile --profile release

# Zephyr:
west build -b <board> -- -DCMAKE_EXPORT_COMPILE_COMMANDS=ON
fw-context index build/compile_commands.json

# PlatformIO:
pio run --target compiledb
fw-context index

# CMake / Make:
bear -- make
# or: bear -- cmake --build build

fw-context index

# 4. Restart your assistant and start asking about your code

For detailed prerequisites, Ollama setup, and AI assistant integration: Installation guide →

Why not just use LSP?

LSP servers (clangd, ccls) are excellent for interactive editing. But they have limitations for AI-assisted exploration:

Use clangd for editing, fw-context for AI-assisted exploration.

Architecture

Data flow

   BUILD                          INDEX                          QUERY
   =====                          =====                          =====
   bear / west / pio    libclang parses each TU          AI assistant calls
   cmake / make         extracts symbols + refs          MCP tools over
        │               generates embeddings            JSON-RPC (stdio)
        ▼                       │                              │
   compile_commands      SQLite db on disk               lookup_symbol(…)
   .json                 ~/.fw-context/index/            search_code(…)
                         │            │                  find_callers(…)
                         ▼            ▼                  explain_symbol(…)
                    symbols + refs   vec0                 get_symbol_context(…)
                    (FTS5 index)   (vector KNN)                 │
                                                          ▼
                                                    AI assistant answers
                                                    your question about
                                                    the code

Components

   CLI (fw-context)            MCP server (fw-context-mcp)          Ollama (optional)
   ================            ===========================          ==================
   fw-context index            exposes 19 tools over               local LLM runtime
   fw-context export           JSON-RPC (stdio)                    HTTP :11434
   fw-context watch                  │                                  │
   fw-context status           search_code ───────────── lookup   smart_search ──▶ translates NL → FTS5 terms
   fw-context reset            lookup_symbol ─────────── prefix   explain_symbol ─▶ explains function
   fw-context init             smart_search ──────────── NL       embeddings ────▶ mxbai-embed-large
   fw-context search           get_file_map ──────────── file structure by kind
                               get_source ────────────── body
                               get_symbol_context ────── body+callers+callees
                               find_callers ──────────── direct callers
                               find_references ───────── all uses
                               find_call_path ────────── BFS in call graph
                               find_all_callers_recursive  transitive callers
                               find_callees_recursive ── transitive callees
                               find_dead_code ────────── never called
                               find_hotspots ─────────── most-called
                               get_active_build ──────── index health
                               reindex_file ──────────── re-parse one file
                               reset_index ───────────── delete + rebuild
                               list_projects ─────────── all indexed projects
                               check_ollama ──────────── verify LLM

Key capabilities

• Fast lookups — FTS5 full-text search, prefix/exact symbol lookup, call-graph traversal
• Natural-language search — "how does the modem connect?" → finds network_registration, modem_attach, … (Ollama, optional)
• Vector search — semantic similarity via sqlite-vec + Ollama embeddings, hybrid FTS5+vector re-ranking
• Graph analytics — call paths, transitive callers/callees, dead code detection, hotspot analysis
• Incremental indexing — only changed files re-parsed; auto-reindex on query detects and fixes staleness
• Offline-first — index is a file on disk at ~/.fw-context/index/. No daemon, no cloud, no network.
• #ifdef-aware — uses real compiler flags; sees exactly what your compiler sees

Supported ecosystems

Works with any build system that produces compile_commands.json:

Subsequent runs are incremental — seconds for a few changed files.

Documentation

Directory layout

~/.fw-context/
├── config.toml              # global defaults
├── .venv/                   # Python virtual environment
│   └── bin/
│       ├── fw-context       # CLI
│       └── fw-context-mcp   # MCP server
└── index/
    └── <project-id>/
        └── index.db         # SQLite + FTS5 + vec0 + refs

your-firmware/
├── .fw-context/
│   └── config.toml          # per-project overrides
└── compile_commands.json