What did Hlido score Toolport?

Toolport scored 84/100 (STEADY) in Hlido's independent, hands-on review.

How much does Toolport cost?

Hlido recorded the pricing model as: Open source. Derived from evidence Hlido holds, not vendor-supplied; verify current prices on the vendor's pricing page.

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.

Toolport

Name: Toolport review
Item: Toolport
Rating: 84
Author: Hlido Editor

Infrastructure · tested 2026-08-09 · re-test due 2026-11-09 · by the Hlido desk, not the vendor

In short: A local MCP gateway that does the two things this layer should do — collapse per-client config and stop tool definitions eating the context window — and it publishes a reproducible benchmark for the second claim.

6 PASS · 1 FAIL of 7 public-surface claims

Visit Toolport → scorecard.json

Quick answer

Toolport scores 84/100 (STEADY) on Hlido’s independent, hands-on test (reviewed 2026-08-09). STEADY (84) for correctly identifying both problems at this layer (config duplication and context cost), answering them with on-demand meta-tools rather than aggregation alone, defending the two real MCP attack patterns Pricing: Open source (free entry point documented).

Hlido Editor · 2026-08-09

Toolport sits between every MCP client on a machine and every MCP server, so a server is configured and authenticated once and then toggled on in Claude, Cursor, VS Code, Windsurf, Codex or Antigravity without restarts. The headline technical claim is context reduction: instead of each server dumping its full tool list into the prompt, Toolport exposes a handful of meta-tools the agent searches on demand. The surface quantifies it — 23,698 tool-definition tokens on one request without, 886 with, and a stated 74–91% reduction across servers 'graded at the same task success' — and, unusually for this category, links a reproducible benchmark rather than asking to be believed. Grading for task success is the right control; a token saving that degrades answers is not a saving. The security posture is the other reason to take this seriously: API keys live in the OS keychain and are injected at runtime rather than sitting in config files, and every tool is fingerprinted so a definition changing after approval (a rug pull) or hidden instructions inside a description (tool poisoning) get flagged locally and by default. Those are the two live attack patterns against MCP, and defending them at the gateway is the correct place. It is MIT-licensed, cross-platform, needs no account or cloud, and reports per-server latency, error rates and an audit trail. The caution is maturity rather than design: 138 GitHub stars, no version or changelog on the captured page, no named production deployments, and — for a component that every tool call now passes through — no stated availability or overhead figures.

Why STEADY

STEADY (84) for correctly identifying both problems at this layer (config duplication and context cost), answering them with on-demand meta-tools rather than aggregation alone, defending the two real MCP attack patterns (rug pulls and tool poisoning) locally and by default, keeping secrets in the OS keychain, and — the part that earns the most credit — publishing a reproducible benchmark graded at equal task success instead of a bare token number. Held below VITAL by early adoption (138 stars), no version or changelog on the surface, and no availability or overhead figures for a component that is now in the path of every tool call.

Public-surface checklist

PASS Homepage loads (required)
PASS Primary value prop (required) — 'All your MCP servers. One port.'
PASS Cta present (required) — 'Download' plus 'Star on GitHub'
PASS Pricing or access — Free and open source (MIT); Teams tier referenced separately
PASS Docs present (required) — Features, Agents, Teams, Security and supported-agent pages
PASS Quantified claim supported — 74–91% token reduction with a linked reproducible benchmark, graded at equal task success
FAIL Third party validation — No named production deployments or independent review on the surface

What we saw

4 screenshots captured by the Hlido engine during the reviewed run (run-fcd8208d3b2f3434-toolport-app). Our own captures — not vendor marketing material.

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

Toolport — run screenshot 2 (page_.png) — page_.png

Toolport — run screenshot 3 (page__features.png) — page__features.png

Toolport — run screenshot 4 (page_clients_.png) — page_clients_.png

What it does well

One gateway, one config: authenticate a server once and toggle it in every agent with no restarts
On-demand meta-tools keep tool definitions out of context — a stated 74–91% token reduction, with a linked reproducible benchmark
Benchmark is graded at equal task success, not on token count alone
Tool fingerprinting flags rug pulls and tool poisoning locally, on by default — the two live MCP attack patterns
API keys held in the OS keychain and injected at runtime, never written to a config file
Per-tool governance: one switch hides every destructive tool from every agent
MIT-licensed, cross-platform, no account and no cloud dependency
Live observability — per-server latency, error rates and an audit trail

What it fails at

Early adoption — 138 GitHub stars and no named production deployments on the surface
No version or changelog published on the captured page
No availability, overhead or latency figures for the gateway itself, which now sits in every tool call
Introduces a single point of failure between every agent and every server
Not tested hands-on by Hlido — the context-reduction and integrity-check claims are the ones that most need measuring

Best for

Developers running several MCP servers across more than one AI client
Long agent loops where tool-definition tokens are consuming meaningful context budget
Anyone who wants rug-pull and tool-poisoning detection without sending anything to a vendor

Not recommended for

Setups with a single MCP server, where the gateway adds a dependency and little benefit
Teams that require a stated availability commitment for an always-in-path component
Anyone who cannot accept a new local single point of failure

Pricing & access

ModelOpen source
Free entry pointYes — a free tier or open-source edition is documented
Pricing findable on the public surfacePASS Free and open source (MIT); Teams tier referenced separately (tested 2026-08-09)

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 2026-08-09.

Related agents

Agent relevance

API CLI MCP Behavioral-testable

Agentic-Commerce Readiness 77/100 · COMMERCE-READY

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

The product IS the agent integration layer: a local MCP gateway every MCP-aware client connects to on one port, exposing searchable meta-tools instead of full tool lists. Agents are the only consumer.

Agent-friendly score: 9/10

Evidence

Free, open-source (MIT) local MCP gateway for Windows, macOS and Linux with no account or cloud — source (2026-08-09) verified
Works with Claude, Cursor, VS Code, Windsurf, Codex and Antigravity — source (2026-08-09) verified
Tool-definition tokens 23,698 without vs 886 with; 74–91% reduction across servers graded at equal task success — source (2026-08-09) verified
A reproducible benchmark is published for the token-reduction claim — source (2026-08-09) verified
Fingerprints every tool and flags rug pulls and tool poisoning, locally and on by default — source (2026-08-09) verified
API keys stored in the OS keychain and injected at runtime — source (2026-08-09) verified
138 GitHub stars at capture — source (2026-08-09) verified
Version, changelog, or named production deployment — source (2026-08-09)

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-2+editorial-narrative-v2 · Methodology version 2026.05 · Next review due 2026-11-09

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/tsouth89-toolport.svg)](https://hlido.eu/check/?agent=tsouth89-toolport)

HTML

<a href="https://hlido.eu/check/?agent=tsouth89-toolport"><img src="https://hlido.eu/badge/tsouth89-toolport.svg" alt="Hlido trust score"></a>