Taste
Wire human expert review and verifiable on-chain attestation into agent workflows before you ship AI-generated work.
Overview
Taste is an MCP server for the Ship phase that routes AI agent work through expert human review and returns on-chain proof that the review happened.
What is this MCP server?
- Remote MCP over streamable HTTP at api.humantaste.app/mcp/
- Positions expert human review specifically for AI agent deliverables
- On-chain proof of human review for auditability and trust signals
- Version 2.0.0 with GitHub source at with0utwhy/taste-mcp
- No local server binary—connect via hosted remote in MCP client config
- Server version 2.0.0 per server metadata
- Single streamable-http remote endpoint published in the MCP manifest
- GitHub repository: github.com/with0utwhy/taste-mcp
What problem does it solve?
Agent output can ship fast but still feel unvetted, with no portable proof a human expert actually reviewed it.
Who is it for?
Indie builders shipping agent-facing products who want a documented human review step and a trust signal beyond automated checks.
Skip if: Teams that only need private internal PR review, or builders who cannot use hosted review services or on-chain attestations.
What do I get? / Deliverables
After you add the Taste remote MCP server, your agent can request expert review and surface on-chain attestations you can cite in launch materials or compliance narratives.
- Expert review outcomes callable from agent sessions
- On-chain proof artifacts associated with reviewed work
- Repeatable review workflow without leaving the MCP tool surface
Recommended MCP Servers
Journey fit
Expert review is a pre-release quality gate—same journey moment as code review and launch prep, not feature implementation. The server’s purpose is third-party review and proof of review, which maps directly to the Ship → review shelf for solo builders validating agent output.
How it compares
Human-review attestation MCP, not an automated code-review skill or static linter.
Common Questions / FAQ
Who is app.humantaste/taste-mcp for?
Solo builders and small teams using AI coding agents who want expert human review plus verifiable on-chain proof before they ship customer-facing work.
When should I use app.humantaste/taste-mcp?
Use it in the Ship phase when agent-generated specs, copy, or features need a human expert sign-off and you want that decision recorded as proof, not just chat history.
How do I add app.humantaste/taste-mcp to my agent?
Add the streamable-http remote https://api.humantaste.app/mcp/ in your MCP client configuration, then follow Taste’s llms.txt and site docs for authentication and tool usage.