Skip to content

depwire/depwire

MCP Developer Tools

A deterministic, symbol‑level dependency graph generator for multi‑language codebases that powers precise What‑If simulations, security scans, and architecture health checks.

TypeScript Latest v1.7.0 · 6d ago Security brief →

Features

  • Deterministic symbol‑level dependency graphs across 16 languages
  • What‑If change simulation with exact impact analysis
  • Security scanning based on the precise graph
  • Architecture health scoring and cross‑language edge detection

Recent releases

View all 39 releases →
No immediate action
v1.7.0 Breaking risk

depwire diff command

No immediate action
v1.6.1 Bug fix

Dead code scoring fix

No immediate action
v1.6.0 New feature

CLI verification command

No immediate action
v1.5.0 New feature

R support

v1.4.0 New feature
⚠ Upgrade required
  • .depwire/claims.jsonl and .depwire/decisions.jsonl should be added to .gitignore
Notable features
  • Pre‑action verification tool `verify_change` providing deterministic safety reports
  • Multi‑agent file claiming tools: `claim_files`, `release_files`, and `get_active_claims`
  • Decision memory tools: `record_decision` and `get_decisions`
Full changelog

What's new

This release adds 6 new MCP tools designed for multi-agent AI coding workflows. Whether you're a solo developer using Cursor, an autonomous agent making code changes, or an orchestrator coordinating multiple agents — these tools give you deterministic verification, coordination, and decision memory.

New MCP Tools (6)

Pre-action verification

  • verify_change — Pass a proposed code change (file + content, or unified diff), get a deterministic safety report: broken imports, new circular dependencies, security findings, health score delta, blast radius, and risk level. Extends "What If" simulation to actual proposed code.

Multi-agent coordination

  • claim_files — Declare which files an agent is about to modify. Prevents parallel agents from stomping each other's work. Returns conflicts if files are already claimed.
  • release_files — Release a previously made claim when work is done.
  • get_active_claims — Query who is currently working on what. Useful for orchestrator agents deciding what to delegate.

Decision memory

  • record_decision — Save a structured decision with reasoning so future agents (or the same agent in a future session) can see what was decided and why.
  • get_decisions — Retrieve past decisions matching a query, file, tag, or time range.

Tool count

17 → 23 MCP tools total.

Storage

The new tools persist state in a .depwire/ directory in your project:

  • .depwire/claims.jsonl — Multi-agent file claims (append-only audit log)
  • .depwire/decisions.jsonl — Agent decision records (append-only audit log)

Recommended addition to your project's .gitignore:

.depwire/claims.jsonl
.depwire/decisions.jsonl

Forward compatibility with AIT

All 6 new tools accept an optional agent_identity_token parameter, reserved for future integration with the AIT (Agent Identity Token) open standard. Tools work without it today; future releases will use it for cryptographic verification of agent actions.

Supported languages (15)

TypeScript · JavaScript · Python · Go · Rust · C · C# · Java · C++ · Kotlin · PHP · Swift · Mojo · Ruby · Dart

Install

npm install -g [email protected]

Or for MCP:

mcp install io.github.atef-ataya/depwire

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

About

Stars
42
Forks
4
Languages
TypeScript HTML JavaScript
Downloads/week
8 ↑17%
NPM Maintainers
2
Contributors
1

Beta — feedback welcome: [email protected]