MeatSpace

zmarten/meatspace

Hooks Claude into a live human reviewer for decisions the model can't or shouldn't make alone. Exposes three MCP tools: provision_api_key (self-service onboarding), get_service_status, and ask_human, which submits a title, optional content (text, markdown, HTML, or image), and 2-4 labeled choices, then long-polls for the human's pick. Built for approval gates before deploys or payments, subjective tie-breaks below confidence thresholds, and taste calls where deterministic checks fall short. The same backend powers a REST API and browser SDK. Runs on meatspace.run with Supabase for persistence and webhook callbacks for async flows.

tools: 3
auth: Secret needed
transport: HTTP
last updated: May 7, 2026

View on GitHub

Install

From registry metadata

$ npx add-mcp https://meatspace.run/api/mcp --name meatspace --header 'Authorization: Bearer ${API_TOKEN}'

Run in your terminal. add-mcp asks which agents to add it to, then asks for each value it needs. Add -g to install it for every project.

Review the command, arguments and environment values before installing. MCP servers run with your local permissions.

Other ways

Claude Code

$ claude mcp add --transport http meatspace https://meatspace.run/api/mcp --header 'Authorization: YOUR_AUTHORIZATION'

Run in your terminal. Add --scope user to make it available in every project.

Tools

Verified live · 3 tools

Verified live against the running server on Nov 6, 2026.

get_service_statusCheck whether MeatSpace is available and when to use a human. Call this when deciding whether to escalate a subjective or high-consequence choice.

Check whether MeatSpace is available and when to use a human. Call this when deciding whether to escalate a subjective or high-consequence choice.

No parameters. Call it with no arguments.

ask_humanPresent content to a human and ask them to choose between options. Use this for subjective judgment, approval, preference, or tie-breaks. Avoid using it for deterministic checks or reversible low-stakes choices. The tool waits briefly for a result, then returns pending if the...14 params

Present content to a human and ask them to choose between options. Use this for subjective judgment, approval, preference, or tie-breaks. Avoid using it for deterministic checks or reversible low-stakes choices. The tool waits briefly for a result, then returns pending if the...

Parameters* required

agent_name*string

Your agent/tool name (max 100 chars)

callback_urlstring

Optional HTTPS webhook URL for async notification

choices*array

2-4 choices for the human to pick from

confidencenumber

Agent confidence between 0 and 1

consequence_of_wrong_choicestring

Why a wrong choice matters (max 500 chars)

contentstring

Content for the human to review (text, markdown, HTML, or image URL). Max 50KB.

content_typestring

How to render the content. Default: textone of text · markdown · html · image

decision_reasonstring

Why the agent is escalating this to a human (max 500 chars)

metadataobject

Optional metadata passed through to webhook

recommended_optionstring

Optional choice id the agent currently recommends

run_idstring

Optional workflow run identifier

timeout_secondsnumber

Request expiry in seconds (default 3600, max 86400)

title*string

Short title for the request (max 200 chars)

trace_idstring

Optional trace identifier

provision_api_keyCreate a MeatSpace API key instantly. No authentication required. Returns a Bearer token for use with the ask_human tool. Max 5 active keys per email address.2 params

Create a MeatSpace API key instantly. No authentication required. Returns a Bearer token for use with the ask_human tool. Max 5 active keys per email address.

Parameters* required

email*string

Owner email address

name*string

Agent or tool name (max 100 chars)

Readme

MeatSpace

Human-in-the-loop service for AI agents. When an agent hits a subjective, high-stakes, or ambiguous decision, MeatSpace routes it to a human who picks one of 2–4 options and returns a structured result.

Live at meatspace.run

What it does

An agent posts a title, optional content (text, markdown, HTML, or an image), and 2–4 labeled choices. A human reviewer is shown the request, picks one, and the API returns the selected id and label. The agent waits via long-poll or webhook.

Typical use cases:

Approval gates before destructive or irreversible actions (deploys, deletes, payments).
Subjective tie-breaks where the model is below its confidence threshold.
Tasteful judgment calls — copy choices, design preferences, ranking ties.
Escalation when an agent has run out of deterministic checks.

Don't use it when the task is deterministic, automatically verifiable, or low-stakes and easily reversible.

Three integration methods

View the full README on GitHub

Details

Repository: github.com/zmarten/meatspace
Website: meatspace.run
Transport: HTTP
Auth: Required
Registry: MCP registry, active
Tools verified: Nov 6, 2026