indxr gives AI agents a persistent, self-updating wiki about your codebase backed by a fast structural index. Architecture decisions, module responsibilities, failure patterns — stuff that normally lives in people's heads — gets written down and stays current.

Features

• Codebase knowledge wiki — persistent wiki with architecture pages, module overviews, failure pattern recording, contradiction tracking, and knowledge compounding. Agents query it, learn from it, and write back to it
• 9 wiki MCP tools — wiki_generate, wiki_search, wiki_read, wiki_contribute, wiki_update, wiki_compound, wiki_suggest_contribution, wiki_record_failure, wiki_status
• Self-updating — wiki automatically stays in sync with code changes via indxr serve --watch --wiki-auto-update
• 26-tool MCP server (3 compound default + 23 granular via --all-tools) — symbol lookup, file summaries, caller tracing, signature search, complexity hotspots, type flow tracking, workspace support
• 27 languages — tree-sitter AST parsing for 8 languages, regex extraction for 19 more
• Token-aware — progressive truncation to fit context windows
• Git structural diffing — declaration-level diffs (+ added, - removed, ~ changed) against any git ref or GitHub PR
• Dependency graphs — file and symbol dependency visualization as DOT, Mermaid, or JSON
• File watching — continuous re-indexing as you edit, via indxr watch or indxr serve --watch
• Monorepo / workspace support — Cargo, npm, Go workspaces; --member to scope to specific members
• One-command agent setup — indxr init configures Claude Code, Cursor, Windsurf, Codex CLI
• Incremental caching — mtime + xxh3 content hashing
• Complexity hotspots — per-function cyclomatic complexity, nesting depth, parameter count; codebase health reports
• Composable filters — path, kind, symbol name, visibility, language

Install

cargo install indxr --features wiki

This installs indxr with the full wiki system. For additional transports:

cargo install indxr --features wiki,http   # + Streamable HTTP transport

Or build from source:

git clone https://github.com/bahdotsh/indxr.git
cd indxr && cargo build --release --features wiki

You can also install without the wiki feature (cargo install indxr) if you only need structural indexing.

Usage

# Wiki — the core workflow
indxr wiki generate                          # generate codebase knowledge wiki
indxr wiki update                            # update wiki after code changes
indxr wiki status                            # check wiki health
indxr wiki compound notes.txt                # compound knowledge from file

# MCP server — live queries + wiki tools for AI agents
indxr serve ./my-project --watch --wiki-auto-update  # recommended: full setup
indxr serve ./my-project                     # start MCP server (structural only)
indxr serve ./my-project --watch             # MCP server with auto-reindex

# Structural indexing
indxr                                        # index cwd → stdout
indxr ./my-project -o INDEX.md               # index project → file
indxr -f json -l rust,python -o index.json   # JSON, filter by language

# Setup
indxr init                                   # set up all agent configs
indxr members                                # list workspace members (monorepo)

Codebase Knowledge Wiki

The structural index tells agents what exists in your codebase. The wiki tells them why. Architecture decisions, module responsibilities, failure patterns — things that usually get lost when someone leaves the team.

indxr wiki generate                          # generate wiki from scratch
indxr wiki update                            # update after code changes
indxr wiki status                            # check wiki health
indxr wiki compound notes.txt                # compound knowledge from file
echo "synthesis" | indxr wiki compound -     # compound from stdin

Wiki pages are stored in .indxr/wiki/ as Markdown with YAML frontmatter. Page types: architecture, module, entity, topic. Pages support [[page-id]] cross-references, contradiction tracking, and failure pattern recording.

How agents use the wiki

1. Generate: wiki_generate → agent plans pages from structural context, writes each via wiki_contribute
2. Query: wiki_search to understand modules and design decisions before reading source
3. Learn: wiki_compound persists synthesized insights after cross-page analysis
4. Record failures: wiki_record_failure so future agents don't repeat mistakes
5. Update: wiki_update identifies stale pages after code changes, agent rewrites via wiki_contribute

Wiki MCP tools (9 tools)

wiki_generate is always listed; the remaining 8 tools appear once a wiki exists. Wiki tools support contradiction tracking and failure pattern recording.

Auto-updating wiki

The MCP server keeps the wiki in sync with your code automatically:

indxr serve --watch --wiki-auto-update

Triggers wiki page updates when source files change. Set ANTHROPIC_API_KEY, OPENAI_API_KEY, or use --exec for a custom LLM backend.

See Wiki docs for full details on page structure, LLM configuration, and disk layout.

Agent Setup

indxr init                    # set up for all agents
indxr init --claude           # Claude Code only
indxr init --cursor           # Cursor only
indxr init --windsurf         # Windsurf only
indxr init --codex            # OpenAI Codex CLI only
indxr init --global           # install globally for all projects
indxr init --global --cursor  # global Cursor only
indxr init --no-rtk           # skip RTK hook setup

Agents don't always pick MCP tools over file reads on their own. indxr init sets up PreToolUse hooks that intercept Read/Bash calls and instruction files that teach the exploration workflow.

MCP Server

JSON-RPC 2.0 over stdin/stdout (or Streamable HTTP with --features http). 3 compound tools listed by default; --all-tools exposes all 26. Wiki tools (9) are available when a wiki exists.

Default tools (3 compound)

Granular tools (23 — requires --all-tools)

Granular tools are always callable even when not listed — --all-tools only controls whether they appear in tools/list.

Wiki tools (9) are also available when a wiki exists. See Wiki section.

In workspace mode, tools gain a member param to scope queries. List tools support compact mode. See MCP Server docs for details.

Output

Default format is Markdown at signatures detail level:

# Codebase Index: my-project

> Generated: 2025-03-23 | Files: 42 | Lines: 8,234
> Languages: Rust (28), Python (10), TypeScript (4)

## Directory Structure
src/
  main.rs
  parser/
    mod.rs
    rust.rs

## src/main.rs

**Language:** Rust | **Size:** 1.2 KB | **Lines:** 45

**Declarations:**
`pub fn main() -> Result<()>`
`pub struct App`

Filtering

indxr --filter-path src/parser              # subtree
indxr --kind function --public-only         # public functions only
indxr --symbol "parse"                      # symbol name search
indxr -l rust,python                        # language filter
indxr --filter-path src/model --kind struct --public-only  # combine

Filters compose. --kind accepts: function, struct, class, trait, enum, interface, module, method, constant, impl, type, namespace, macro.

Git Structural Diffing

indxr --since main
indxr --since v1.0.0
indxr --since HEAD~5
indxr diff --pr 42                           # diff against a GitHub PR's base branch

## Modified Files

### src/parser/mod.rs
+ `pub fn new_parser() -> Parser`
- `fn old_helper()`
~ `fn process(x: i32)` → `fn process(x: i32, y: i32)`

Markers: + added, - removed, ~ signature changed.

Complexity Hotspots

indxr --hotspots                             # top 30 most complex functions
indxr --hotspots --filter-path src/parser    # scoped to a directory

Shows cyclomatic complexity, max nesting depth, parameter count, body lines, and composite score per function. Tree-sitter parsed languages only.

MCP tools: get_hotspots, get_health, get_type_flow.

Dependency Graph

indxr --graph dot                            # file-level DOT graph
indxr --graph mermaid                        # file-level Mermaid diagram
indxr --graph json                           # JSON graph
indxr --graph dot --graph-level symbol       # symbol-level graph
indxr --graph mermaid --filter-path src/mcp  # scoped to a directory
indxr --graph dot --graph-depth 2            # limit to 2 hops

Token Budget

indxr --max-tokens 4000

Truncation order: doc comments → private declarations → children → least-important files. Directory tree and public API surface are preserved first.

Languages

8 tree-sitter (full AST) + 19 regex (structural extraction):

Detection is by file extension. Full details: docs/languages.md

Performance

Parallel parsing via rayon. Incremental caching via mtime + xxh3.

Documentation

Contributing

Contributions welcome — feel free to open an issue or submit a PR.

License

MIT