CICADA

Code Intelligence: Contextual Analysis, Discovery, and Attribution

Give your AI assistant structured access to your Elixir codebase.

Quick Install · Security · Developers · AI Assistants · Docs

What's New in 0.3

• Dependency analysis tools - New get_module_dependencies and get_function_dependencies with transitive depth support
• Expanded editor support - Added Gemini CLI, Codex, and OpenCode (now 6 editors supported)
• Watch mode - Automatic reindexing with cicada watch or --watch flag
• Better discoverability - Smart error suggestions, inline PR context, staleness warnings
• Wildcard & OR patterns - Search with MyApp.*, create*|update* across modules and functions
• Breaking: Removed CICADA_REPO_PATH environment variable

Why CICADA?

Traditional AI assistants treat your repo like a pile of text. That leads to:

• Token waste: blind grep dumps that burn 3k+ tokens per question.
• Hallucinated edits: aliases/imports hide call sites, so refactors miss real usages.
• No historical context: design intent and PR trade-offs never make it into the prompt.

CICADA is an MCP server that gives assistants AST-level knowledge:

• Module + function definitions with signatures, specs, docs, owning files.
• Complete call-site tracking (aliases, imports, dynamic references).
• Semantic/keyword search so you can ask for "authentication" even if it's called verify_credentials/2.
• Git + PR attribution to surface why code exists.
• Dead-code detection and module dependency views for safe refactors.

Result: in our comparison, the same question dropped from 3,127 tokens / 52.8s to 550 tokens / 35s with correct answers.

Quick Install

# 1. Install uv (if needed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# 2. Install CICADA
uv tool install cicada-mcp

# 3. Index your Elixir project
cd /path/to/project
cicada claude   # or: cicada cursor, cicada vs, cicada gemini, cicada codex, cicada opencode

# 4. Restart your editor

uvx --from cicada-mcp cicada claude   # or cursor, vs

Runs CICADA on demand (slower after the first run, but zero install).

Available commands after installation:

• cicada [claude|cursor|vs|gemini|codex|opencode] - One-command setup per project
• cicada-mcp - MCP server (auto-started by editor)
• cicada watch - Watch for file changes and automatically reindex
• cicada index - Re-index code with custom options (-f/--force + --fast/--regular/--max, --watch)
• cicada index-pr - Index pull requests for PR attribution
• cicada find-dead-code - Find potentially unused functions

Ask your assistant:

"Show me the functions in MyApp.User"
"Where is authenticate/2 called?"
"Find code related to API authentication"

Privacy & Security

• 100% local: parsing + indexing happen on your machine; no cloud uploads.
• No telemetry: CICADA doesn't collect usage or phone home.
• Read-only tools: MCP endpoints only read the index; they can't change your repo.
• Optional GitHub access: PR features rely on gh and your existing OAuth token.
• Data layout:
  Copy

  ~/.cicada/projects/<repo_hash>/
  ├─ index.json      # modules, functions, call sites, metadata
  ├─ config.yaml     # indexing options + keyword tier
  ├─ hashes.json     # incremental indexing cache
  └─ pr_index.json   # optional PR metadata + reviews
  

  Your repo only gains an editor config (.mcp.json, .cursor/mcp.json, .vscode/settings.json, .gemini/mcp.json, .codex/mcp.json, or .opencode.json).

For Developers

Wire CICADA into your editor once, and every assistant session inherits the context.

Install & Configure

cd /path/to/project
cicada claude   # or cicada cursor / cicada vs / cicada gemini / cicada codex / cicada opencode

This command:

1. Parses every .ex/.exs file with tree-sitter.
2. Builds the index in ~/.cicada/projects/<hash>/.
3. Creates the correct MCP config for your editor.
4. Configures .gitattributes so git can track functions through refactors.

Enable PR Attribution (optional)

brew install gh    # or apt install gh
gh auth login
cicada index-pr .     # incremental
cicada index-pr . --clean   # full rebuild

Unlocks questions like "Which PR introduced line 42?" or "What did reviewers say about billing.ex?"

Automatic Re-indexing with Watch Mode

Enable automatic reindexing when files change by starting the MCP server with the --watch flag:

For Claude Code (.mcp.json):

{
  "mcpServers": {
    "cicada": {
      "command": "cicada-mcp",
      "args": ["--watch"],
      "env": {
        "CICADA_CONFIG_DIR": "/home/user/.cicada/projects/<hash>"
      }
    }
  }
}

For Cursor (.cursor/mcp.json):

{
  "mcpServers": {
    "cicada": {
      "command": "cicada-mcp",
      "args": ["--watch"]
    }
  }
}

Or use with the server subcommand:

cicada server --watch
cicada server --watch --fast    # Use fast tier for reindexing
cicada server --watch --max     # Use max tier for reindexing

When watch mode is enabled:

• A separate process monitors .ex and .exs files for changes
• Changes are automatically reindexed (incremental, fast)
• 2-second debounce prevents excessive reindexing during rapid edits
• The watch process stops automatically when the MCP server stops
• Excluded directories: deps, _build, node_modules, .git, assets, priv

CLI Cheat Sheet

Troubleshooting

Run the indexer first:

cicada index /path/to/project

Ensure indexing completed successfully. Check for ~/.cicada/projects/<hash>/index.json.

Use the exact module name as it appears in code (e.g., MyApp.User, not User).

If module was recently added, re-index:

cicada index .

Troubleshooting checklist:

1. Verify configuration file exists:

   bashCopy

   # For Claude Code
   ls -la .mcp.json
   
   # For Cursor
   ls -la .cursor/mcp.json
   
   # For VS Code
   ls -la .vscode/settings.json
   

2. Check paths are absolute:

   bashCopy

   cat .mcp.json
   # Should contain: /absolute/path/to/project
   # Not: ./project or ../project
   

3. Ensure index exists:

   bashCopy

   ls -la ~/.cicada/projects/
   # Should show directory for your project
   

4. Restart editor completely (not just reload window)

5. Check editor MCP logs:

   • Claude Code: Console output
   • Cursor: Settings → MCP → View Logs
   • VS Code: Output panel → MCP

Setup GitHub CLI:

# Install GitHub CLI
brew install gh  # macOS
sudo apt install gh  # Ubuntu
# or visit https://cli.github.com/

# Authenticate
gh auth login

# Index PRs
cicada index-pr .

Common issues:

• "No PR index found" → Run cicada index-pr .
• "Not a GitHub repository" → Ensure repo has GitHub remote
• Slow indexing → First-time indexing fetches all PRs; subsequent runs are incremental
• Rate limiting → GitHub API has rate limits; wait and retry if you hit limits

Force rebuild:

cicada index-pr . --clean

Error: "Keyword search not available"

Cause: Index was built without keyword extraction.

Solution:

# Re-index with keyword extraction
cicada index --regular .  # or --fast or --max

Verify:

cat ~/.cicada/projects/<hash>/config.yaml
# Should show keyword_extraction: enabled

More detail: docs/PR_INDEXING.md, docs/08-INCREMENTAL_INDEXING.md.

For AI Assistants

CICADA ships eleven focused MCP tools. Use the decision table to pick the right one:

🧭 Which Tool Should You Use?

Want to see these tools in action? Check out Complete Workflow Examples with pro tips and real-world scenarios.

Core Search Tools

search_module - Find modules and view all their functions

• Search by module name or file path with wildcards (MyApp.*, *User*) and OR patterns (MyApp.User|MyApp.Post)
• View function signatures with type specs
• Filter public/private functions
• Output in Markdown or JSON

search_function - Locate function definitions and track usage

• Search by function name, arity, or module path with wildcards (create*, *_user) and OR patterns (create*|update*)
• See where functions are called with line numbers
• View actual code usage examples
• Filter for test files only

search_module_usage - Track module dependencies

• Find all aliases and imports
• See all function calls to a module
• Understand module relationships
• Map dependencies across codebase

Git History & Attribution Tools

find_pr_for_line - Identify which PR introduced any line of code

• Line-level PR attribution via git blame
• Author and commit information
• Direct links to GitHub PRs
• Requires: GitHub CLI + PR index

get_file_pr_history - View complete PR history for a file

• All PRs that modified the file
• PR descriptions and metadata
• Code review comments with line numbers
• Requires: GitHub CLI + PR index

get_commit_history - Track file and function evolution over time

• Complete commit history for files
• Function-level tracking (follows refactors)
• Creation and modification timeline
• Requires: .gitattributes configuration

get_blame - Show line-by-line code ownership

• Grouped authorship display
• Commit details for each author
• Code snippets with context

Advanced Features

search_by_features (Beta) - Search code by concepts and features

• 🎯 Perfect for: "I don't know the exact name" - Search by what code does, not what it's called
• Find code related to concepts like "authentication", "api key storage", "email validation"
• Wildcard pattern matching (create*, *_user, validate_*)
• Filter results by type: modules only, functions only, or all
• AI-powered keyword extraction from documentation
• Relevance scoring to surface the most relevant results
• Requires: Index built with keyword extraction (--fast, --regular, or --max)

When to use: You know what you're looking for conceptually but not the exact module/function names. Instead of guessing names with search_function, describe what the code does!

find_dead_code - Identify potentially unused functions

• Three confidence levels (high, medium, low)
• Smart detection of callbacks and behaviors
• Recognition of dynamic call patterns
• Module-level grouping with line numbers
• Excludes test files and @impl functions

Detailed parameters + output formats: MCP_TOOLS_REFERENCE.md.

Token-Friendly Responses

All tools return structured Markdown/JSON snippets (signatures, call sites, PR metadata) instead of full files, keeping prompts lean.

Learn by Doing (5–10 min each)

1. Safe Refactor Checklist

1. search_function → "Where is create_user/2 called?"
2. search_module_usage → "Which modules alias MyApp.User?"
3. search_function with test_only:true to confirm test coverage.
4. get_file_pr_history → "Show PRs that modified lib/my_app/user.ex."

2. Untangle Legacy Intent

1. search_module to skim the API.
2. get_file_pr_history for design discussions/reviews.
3. get_commit_history on the hot function.
4. get_blame on confusing lines to ping the right author.

3. Cleanup Sprint

1. find_dead_code --min-confidence high for candidates.
2. For each, search_function to double-check dynamic usage.
3. find_pr_for_line to ensure it isn't waiting on an unfinished feature.
4. Remove or deprecate confidently.

For full walkthroughs see docs/17-WORKFLOW_EXAMPLES.md and docs/12-TOOL_DISCOVERABILITY_TASKS.md.

Documentation

• CHANGELOG.md – release notes.
• docs/01-KEYWORD_EXTRACTION_ANALYSIS.md – semantic search internals.
• docs/09-PR_INDEXING.md – GitHub integration details.
• docs/16-MCP_TOOL_CALL_BENCHMARKING.md – token/time benchmarks.

Roadmap

Contributing

git clone https://github.com/wende/cicada.git
cd cicada
uv sync
pytest

Before submitting a PR:

• Run black cicada tests
• Ensure tests + coverage pass (pytest --cov=cicada --cov-report=term-missing)
• Update docs if behaviour changes

We welcome issues/PRs for:

• New language grammars
• Tool output improvements
• Better onboarding docs and tutorials

License

MIT – see LICENSE.

Stop letting your AI search blindly. Give it CICADA.

Get Started · Report Issues