MCP.Market

Toolfactory MCP server

by GoatInAHat·io.github.GoatInAHat/toolfactory·v0.4.6
No reviews yet

Build an agent tool once; ship it as Agent Skills, MCP servers, Agent Plugins / Claude / Codex

A92/100grade A
Write a reviewInstall
What users say
No reviews yet
Be the first
Safety scan
A92/100

full report

Adoption
Growing

0 stars636 downloads/wk

Reviews

Write one

Nobody has reviewed Toolfactory yet.

If you have run it, two minutes of your experience saves the next person an afternoon.

Toolfactory tools (22, 4 write)

write = sends, deletes, buys or posts

Read from the package source without running it. The installed server may list more.

  • adopt

    Stop regenerating one file; it becomes the author's (recorded as manual in the lock).

  • bootstrap-repo

    Prepare the GitHub repository from the local .env: the `live-tests` environment with its required reviewers and the sensitive config keys inside it, the release registries' tokens at repository scope, GitHub Pages with source = Actions, and npm's trusted publisher. Values go to `gh` on stdin and are never returned.

  • build

    Generate every selected surface in-tree from the identity file and the operation snapshot, and refresh the lock.

  • check

    Fail if the operation snapshot or any generated file drifted from the code (the CI drift gate).

  • coverage

    The operation × surface verdict matrix: native, bridged, degraded, or excluded, with reasons.

  • doctor

    Report which upstream CLIs this machine can delegate to (git, gh, npm, uv, claude, openclaw, clawhub, hermes, uvx, agentskills, MCP Inspector, docker).

  • echo

    Current operation

  • eject

    Adopt every file a surface owns, so the author takes it over entirely.

  • gatewrite action

    Run the gate here, in order: build, the drift check, every selected surface's upstream validator, the author's checks and tests, and the credential-free host end-to-end. The same step list the generated ci.yml renders, so a project with no CI has the identical gate.

  • initwrite action

    Create a new tool: dev.toolfactory/tool.json, the authored identity file, the kernel scaffold for the chosen language, and the first build of every selected surface.

  • introspect

    Spawn the kernel MCP server, list its tools, and snapshot them to dev.toolfactory/ops.json.

  • new_operation

    Updated operation

  • notifywrite action

    Post a message into the live conversation.

  • package

    Build every release asset into dist/release/ — npm tarball, Python distributions, OpenClaw plugin tarball, plugin bundle zip, web build, coverage — by the same steps the release workflow's package job runs. Publishing stays a CI concern.

  • review-prompt

    Evaluate a prompt draft against the codex-prompt-standard rubric: anatomy sections, convention checks, and concrete directives. Draft, run this, fix what it flags, and re-run until it passes.

  • screenshot

    Capture the current page.

  • secrets-usage

    Every credential this project's surfaces need — the tool's own sensitive config keys and the release registries' tokens — with where each one is set, whether it is present locally and on GitHub, and (check) whether the registry accepts it. Never a value.

  • shoot

    Take a screenshot of the current tab.

  • unadopt

    Return an adopted file to toolfactory and regenerate it.

  • unpublish

    Retract what a deselected surface used to publish. Git is the ledger: the previous tag's dev.toolfactory/tool.json says what was selected then, and every registry row that lost its surface is checked for the version that tag published and then retracted with the registry's own CLI — or reported with the exact page, where there is no API.

  • validatewrite action

    Run each selected surface's own upstream validator (agentskills, claude plugin validate, MCP Inspector, openclaw, hermes, npm pack, uv build).

  • web

    Open the generated operations page.

Public scan report

scanner v0.1.10 · 2026-10-15 · same rubric, same numbers if you re-run it

no findings
  • Code scan262 source files scanned25/25
  • –Live reliabilityno gateway calls yet and no remote to proben/a
  • –Tool poisoningtools not inspected (local package is not executed); not countedn/a
  • Auth qualitylocal package, no credentials required12/15
  • Maintenancelast push 4 days ago15/15
  • Maintainer identityregistry namespace matches repository owner; GitHub account older than a year8/10
Overall 92/100. Components that don't apply are left out of the denominator. Any critical finding is an F.RubricAppeal a findingJSON

How Toolfactory compares

Against every graded AI and agents MCP server, from our own scans and uptime checks. Updated after each scan run.

Safety score
92/100
higher than 90% of 4,460 AI and agents servers
Tools
22
the typical AI and agents server has 6
Tools that change things
18%
typical: 0%. Fewer write tools means less an agent can break
Downloads a week
636
more than 89% of 1,538 AI and agents servers with a package

What its tools let an agent do: 17 read · 1 create · 3 change · 1 send or post

Release history: 3 versions in the registry since 2026-09-16, latest 2026-10-11, typically one every 16 days.

Install Toolfactory in Claude Code, Cursor or VS Code

Runs npx -y toolfactory on your machine. Read the scan report first; the gateway never runs local packages.

claude mcp add toolfactory -- npx -y toolfactory
Add to Cursor

What the publisher says

From the Toolfactory repository's README, as published. We do not edit it. Read it on GitHub

toolfactory

Build an agent tool once; ship it as an Agent Skill, an Agent Plugins bundle, a Claude Code, Codex or Cursor plugin, an MCP server, a CLI, an npm or PyPI package, an MCP Registry entry, an OpenClaw or Hermes native plugin, and a web page, from one operation module and one config file.

toolfactory is a scaffolder and a keeper-in-sync, not a runtime. It writes each surface in the shape that surface's own tooling would have written, proves it with that surface's own validator, and records every generated file in a lock so drift fails CI. Delete toolfactory from a repo it generated and every surface still installs, builds and publishes.

<!-- tf:install -->

Install

MCP endpoint on one port and opens it; over MCP or a skill, the web operation does the same and returns the URL.

  • Agent Skill — npx skills add GoatInAHat/toolfactory
  • MCP server — npx -y toolfactory mcp
  • Claude Desktop extension — download toolfactory.mcpb from the GitHub Release and double-click to install
  • Claude Code plugin — claude plugin marketplace add GoatInAHat/toolfactory, then claude plugin install toolfactory@toolfactory
  • Codex plugin — codex plugin marketplace add GoatInAHat/toolfactory, then codex plugin add toolfactory@toolfactory
  • Gemini CLI extension — gemini extensions install https://github.com/GoatInAHat/toolfactory
  • OpenClaw plugin — openclaw plugins install --link hosts/openclaw from a checkout
  • Hermes plugin — hermes plugins install https://github.com/GoatInAHat/toolfactory#hosts/hermes/toolfactory_hermes
  • DSH plugin (experimental) — dsh plugin --profile add ./hosts/dsh from a checkout, or the release tarball toolfactory-dsh-0.3.1.tgz
  • Web app — npx -y toolfactory mcp --http --open serves the operations page beside the
  • npm package — npm install toolfactory

<!-- /tf:install -->

Quickstart

mkdir hello && cd hello
npx toolfactory init --name hello --binding typescript --surfaces skill,agent-plugins,claude,mcp,cli,npm

init writes plugin.json (identity), dev.toolfactory/tool.json (surfaces, binding, config), the kernel scaffold, the .agents/ agent-config canon (skills, MCP servers, sync.py, setup and the per-harness hook carriers, so the tool is developable in any harness), and the first build. It also does what a human would do next: git init and a first commit, then bash .agents/setup, which renders the harness adapters, installs the git hooks that keep them in sync, and installs the dependencies. Its nextSteps end with the one reload line of the harness it is running inside, because reload is the one part a repository cannot automate.

Add --repo / and it creates that GitHub repository through gh and pushes to it — private unless you pass --public — with one topic per selected surface, and prepares the live-tests environment when a .env is there. --dryRun prints the gh invocations instead. None of it is required: with no GitHub, plain git or no git at all, everything below still works.

Instruction-only Codex plugin

Shortened. The full README is on GitHub.

Nothing above is checked by us. What we check is on the safety report.

Grade history

  • 2026-09-19restoreC → Ascore 92

Toolfactory: common questions

Is Toolfactory MCP server safe?
Yes, by our scan: it is graded A (92/100). Read the Toolfactory safety report
How do I install Toolfactory?
It runs on your machine. Copy the Claude Code, Cursor, VS Code or Claude Desktop config from the install section.
Does Toolfactory need an API key?
Not as far as the registry entry and our scan can tell: no credentials are declared or required.
Is Toolfactory maintained?
The last commit was 24 days ago (2026-10-11). The latest release is v0.4.6.
What can I use instead of Toolfactory?
Servers from other publishers that do the same job: Pm Claude Skills MCP server, SuperCompress MCP server and Watch Skill MCP server.
How does Toolfactory compare with other AI and agents MCP servers?
safety score 92/100 (higher than 90% of 4,460 AI and agents servers); tools 22 (the typical AI and agents server has 6); tools that change things 18% (typical: 0%. Fewer write tools means less an agent can break). Full comparison.

Alternatives to Toolfactory

Same job from other publishers: the closest match first, then the best rated.

  • Pm Claude Skills
    1310 professional Agent Skills + workflow recipes — searchable & fetchable over MCP.
    A
  • SuperCompress
    Compress coding-agent context ~64%. Hosted Neural Keep MCP for Cursor/Claude/Codex.
    B
  • Watch Skill
    Watch video and live sessions, keep timestamped evidence, and verify an agent's own work.
    B
  • mcptoon
    MCP tools + agent skills in one zero-dependency CLI: 71,929 -> 581 tokens (-99.2%, measured).
    A
  • OpenWorkProof
    Verifiable agent evidence; opt-in Codex guard and reference SOC alert review.
    A