What did Hlido score Depwire?

Depwire scored 79/100 (STEADY) in Hlido's independent, hands-on review.

Does any vendor pay Hlido for placement?

No. Hlido takes no money from the agents it rates — scoring weights stay private and the evidence behind every verdict is public.

Depwire

Name: Depwire review
Item: Depwire
Rating: 79
Author: Hlido Editor

Coding · tested 2026-08-05 · re-test due 2026-11-05 · by the Hlido desk, not the vendor

In short: The most substantiated tool in this batch — a deterministic dependency graph served to AI coders over MCP, with real benchmarks and an unusually honest anti-RAG pitch.

5 PASS · 0 FAIL of 5 public-surface claims

Visit Depwire → scorecard.json

Quick answer

Depwire scores 79/100 (STEADY) on Hlido’s independent, hands-on test (reviewed 2026-08-05). STEADY (79) — the strongest public surface in this wave: a shipped v1.8.2 with a real distribution footprint (npm, VSCode Marketplace, GitHub Action, cloud app), 23 documented MCP tools, and a benchmark table across elev Pricing: Open source · Free tier · Subscription (free entry point documented).

Hlido Editor · 2026-08-05

Depwire is a context layer that sits between an AI coding assistant and a codebase: tree-sitter parses every file, graphology builds a symbol-level dependency graph, and 23 MCP tools let Claude, Cursor, Aider or any MCP client query exact facts instead of guessing from embeddings. The positioning is refreshingly precise — "this is not RAG", "Depwire is the GPS, your AI is the driver" — and it declines to claim it replaces the assistant. The public surface is materially better evidenced than most agent tooling we see: a benchmark table across eleven real projects in eleven languages (Hono, Excalidraw, FastAPI, ASP.NET Docs at 5,655 files, nlohmann/json, Spring PetClinic) with file/symbol counts and parse times, a shipped v1.8.2, a VSCode extension, a GitHub Action, a hosted cloud option, and stewardship of an open CC0 spec (DAT) drafted against a NIST concept paper. Feature depth is genuinely broad — what-if simulation, pre-action verification, security scanning with graph-aware severity, structural diff between commits, dead-code detection, multi-agent file claiming. The caveats are the ordinary ones for a young tool: every performance number (40% fewer tool calls, 56% fewer file reads, 100% deterministic accuracy) is vendor-supplied and untested by us; the licence is BSL 1.1 rather than open source until it converts to Apache-2.0 in February 2029; Pro features sit behind $9.99/month; and it appears to be a single-vendor project (ATEF ATAYA LLC). We did not run it. STEADY reflects an unusually well-evidenced public surface, not a measured performance claim.

Why STEADY

STEADY (79) — the strongest public surface in this wave: a shipped v1.8.2 with a real distribution footprint (npm, VSCode Marketplace, GitHub Action, cloud app), 23 documented MCP tools, and a benchmark table across eleven named real-world repositories in eleven languages, which is more falsifiable evidence than nearly any agent tool publishes. Held below VITAL because every efficiency and accuracy figure is vendor-supplied and unverified by us, the licence is BSL 1.1 (not open source until 2029), and we reviewed the public surface only.

Public-surface checklist

PASS Homepage loads (required)
PASS Primary value prop (required) — 'Your AI doesn't know your architecture. Depwire does.'
PASS Cta present (required) — 'Install in 10 seconds' / npm install -g depwire-cli
PASS Pricing or access — Free CLI; Pro features $9.99/month; cloud app available
PASS Evidence or demo — Benchmark table across 11 named real projects; no third-party validation

What we saw

4 screenshots captured by the Hlido engine during the reviewed run (run-694c1c5f43c73d31-depwire-dev). Our own captures — not vendor marketing material.

Depwire — run screenshot 1 (home.png) — home.png

Depwire — run screenshot 2 (page_install.png) — page_install.png

Depwire — run screenshot 4 (pageprivacy.png) — pageprivacy.png

What it does well

Deterministic tree-sitter/AST approach with an explicit, well-argued case against embedding-based retrieval
23 MCP tools — genuinely agent-first, not a human product with an API bolted on
Benchmarks published against eleven named real projects with file counts, symbol counts and parse times
Broad, coherent feature set: impact analysis, what-if simulation, pre-action verification, security scan, structural diff, dead-code detection
Multiple real distribution channels shipped (npm, VSCode extension, GitHub Action, hosted cloud app)
Runs locally by default — code never leaves the machine, and the security scanner needs no API key
Stewards an open CC0 spec (DAT) for signing AI agent actions, drafted against a NIST concept paper

What it fails at

Every headline metric (40% fewer tool calls, 56% fewer file reads, 100% accuracy) is vendor-supplied and unverified by Hlido
BSL 1.1 licence — not open source until it converts to Apache-2.0 in Feb 2029, which blocks some commercial adoption
Apparently a single-vendor, single-author project (ATEF ATAYA LLC) — bus-factor and continuity risk
No named customers, adoption numbers, or third-party validation on the public surface
Pro features and the cloud dashboard are paid ($9.99/month) with limited detail on what the free tier omits
Not tested hands-on — graph accuracy on a large mixed-language monorepo is exactly the claim that needs measuring

Best for

Developers using Claude Code, Cursor, Cline or Aider on a codebase large enough that context retrieval is the bottleneck
Teams doing risky refactors who want blast-radius analysis before an AI edits anything
Mixed-language monorepos where cross-language dependency edges are invisible to text search
Anyone who wants local-only code analysis with no cloud parsing

Not recommended for

Organisations whose licence policy excludes BSL-1.1 dependencies
Buyers who need audited performance figures rather than vendor benchmarks
Small single-file projects where the dependency graph adds no signal
Teams needing vendor-scale support guarantees or an SLA

Pricing & access

ModelOpen source · Free tier · Subscription
Free entry pointYes — a free tier or open-source edition is documented
Price points we recorded…until it converts to Apache-2.0 in February 2029; Pro features sit behind $9.99/month; and it appears to be a single-vendor project (ATEF ATAYA LLC). We did not run…
Pricing findable on the public surfacePASS Free CLI; Pro features $9.99/month; cloud app available (tested 2026-08-05)

Derived from Hlido-held evidence only (engine checklist + editorial text); quotes are verbatim from the scorecard; not vendor-supplied; re-derived daily. Verify current prices on the vendor's pricing page. Last verified 2026-08-05.

Compared to

Context7

Agent relevance

API CLI MCP SDK Behavioral-testable

Agentic-Commerce Readiness 85/100 · COMMERCE-READY

Independent readiness for agent delegation & transaction. How it’s scored · check live

MCP-native by design: `npx -y depwire-cli mcp` exposes 23 tools to any MCP client, with a hosted endpoint via app.depwire.dev and a programmatic SDK at depwire-cli/sdk. This is a component other agents drive, not a destination product — the strongest agent-integration story in this wave.

Agent-friendly score: 9/10

Evidence

23 MCP tools exposed to AI assistants, with a documented tool-by-tool table — source (2026-08-05) verified
Deterministic tree-sitter AST parsing across 17 languages, not embedding-based retrieval — source (2026-08-05) verified
Benchmarks across eleven named real projects (Hono 352 files/6,245 symbols; ASP.NET Docs 5,655 files/42,827 symbols) — source (2026-08-05)
40% fewer tool calls / 56% fewer file reads / 100% deterministic accuracy — source (2026-08-05)
Shipped v1.8.2 under BSL 1.1, converting to Apache-2.0 on 2029-02-25 — source (2026-08-05) verified
Steward of the CC0 Depwire Action Token (DAT) spec for signing AI agent actions — source (2026-08-05) verified

scorecard.json · transparency passport · registry · methodology

More: compare agents · best of · developer tools · incident registry

Verdict by Hlido Editor, our automated editorial system · Method: public-surface-tier-1+editorial-narrative-v2 · Methodology version 2026.05 · Next review due 2026-11-05

How this page was produced. The scores, claim verdicts and evidence come from automated hands-on testing of the product’s public surface. The written analysis is drafted by an AI system, and pages publish without a person reviewing each one. Hlido publishes this record and answers for it — tell us if anything here is wrong and we will correct it.

Embed this trust badge

Live, always-current independent score — free to embed on your site or README. No vendor pays for placement.

Markdown

[![Hlido trust score](https://hlido.eu/badge/depwire-depwire.svg)](https://hlido.eu/check/?agent=depwire-depwire)

HTML

<a href="https://hlido.eu/check/?agent=depwire-depwire"><img src="https://hlido.eu/badge/depwire-depwire.svg" alt="Hlido trust score"></a>