{ "policies": { "deny_new_cycles": true, "max_public_api_growth": 5, "max_coupling_increase": 3 } } ``` Policies are checked automatically by `ariadne verify`. They compare current metrics against the saved baseline snapshot — run `ariadne metrics --snapshot` to establish the baseline. ## CLI commands | Command | Description | |---------|-------------| | `setup [path] [--quick]` | One-command setup: init + index + generate hooks & MCP config. `--quick` skips SCIP. | | `init [path]` | Create `.ariadne/` directory and SQLite database | | `index [path] [--structural-only]` | Full index: structural imports + SCIP analysis. `--structural-only` skips SCIP. | | `update [--full] [--reingest] [--files path...]` | Incremental update: structural imports + auto SCIP when dirty source files exist. `--files` processes only the specified files (used by hooks for fast single-file updates). `--full` forces SCIP even when clean. `--reingest` forces SCIP re-ingestion even when file contents haven't changed (needed after package renames where symbol IDs change). | | `query search ` | Fuzzy search symbols by name | | `query def ` | Get definition location, docs, and code snippet | | `query refs ` | Find all references across the codebase | | `query impact ... [--details]` | Changed files → impacted symbols → dependent files → recommended tests. `--details` adds symbol defs, docs, and up to 3 reference snippets per symbol. | | `query call-graph [--depth N]` | Approximate call graph: callers and callees at the given depth (default 1) | | `query module-graph [--mode] [--path] [--format]` | File dependency graph (imports, semantic, or hybrid) | | `impact ... [--max-depth N] [--call-graph]` | Transitive impact analysis with risk scoring. BFS across symbol graph up to N depths (default 5). `--call-graph` includes caller traversal. | | `metrics [--snapshot] [--diff]` | Structural metrics: coupling (Ca/Ce/I), dependency cycles, API surface. `--snapshot` saves baseline, `--diff` compares against it. | | `context ... [--depth N] [--mode]` | Compile agent-friendly context for files: symbols, imports, importers, semantic refs. Modes: `compact` (default), `full`, `symbols-only`. | | `preflight ... [--depth N]` | Pre-edit blast radius analysis: symbols defined in files, their call sites across the codebase, boundary violations, and risk assessment. | | `ci [--base branch] [--markdown] [--root path]` | CI command: runs impact + verify + metrics on changed files vs base branch. Outputs JSON (default) or `--markdown` for PR comments. | | `verify` | Run gatekeeper checks including policy enforcement (exit 0 = OK, exit 1 = FAIL) | | `status` | Index stats (file count, symbol count, edge count, dirty count) | | `dirty mark ` | Mark a file as needing re-index | | `ledger log ` | Append event to execution ledger | | `doctor [path]` | Check prerequisites (Bun, Node, SCIP indexers for all languages) and warn about stale legacy directories | ## MCP server The MCP server exposes 14 read-only tools for AI agents: | Tool | Description | |------|-------------| | `ariadne.search_symbol` | Fuzzy search symbols by name | | `ariadne.get_def` | Get symbol definition with docs and code snippet | | `ariadne.find_refs` | Find all references to a symbol (optionally scoped) | | `ariadne.impact` | Blast radius analysis for changed files (with optional `details` for symbol defs and key refs) | | `ariadne.transitive_impact` | Full transitive impact with risk scoring, public API break detection, and affected package mapping | | `ariadne.metrics` | Structural metrics: module coupling (Ca/Ce/instability), dependency cycles, API surface per package | | `ariadne.cycles` | Dependency cycle detection using Tarjan's SCC algorithm | | `ariadne.module_graph` | File dependency graph (imports/semantic/hybrid, json/dot/mermaid) | | `ariadne.symbol_graph` | Dependency subgraph centered on a specific symbol | | `ariadne.file_symbols` | List all symbols defined in a file | | `ariadne.status` | Index stats: files, symbols, dirty count, timestamps | | `ariadne.call_graph` | Approximate call graph: callers and callees for a symbol | | `ariadne.plan_context` | Agent-friendly file context: symbols, imports, importers, semantic refs via BFS graph traversal | | `ariadne.preflight` | Pre-edit blast radius: symbols in target files, call sites, boundary risks, and blast radius summary | ### Configuration `ariadne setup` generates this automatically. To configure manually, add `.mcp.json` to your project root: ```json { "mcpServers": { "ariadne": { "command": "bun", "args": ["run", "/path/to/ariadne/packages/mcp/src/index.ts"], "env": { "ARIADNE_ROOT": "." } } } }

Ariadne

Installation