Description
Manufact provides a robust platform for developers to deploy, test, and ship their Multi-Client Protocol (MCP) plugins to various AI marketplaces, including ChatGPT and Claude. It supports MCP servers built with any framework or language, offering build presets for TypeScript and Python, including `mcp-use` and `FastMCP`, alongside a Dockerfile option for other environments.
Read More
The platform streamlines the deployment process, allowing for server deployment in under 60 seconds from a Git push. Developers can connect their GitHub App for automatic deployments on every push or use the CLI. Manufact also offers preview deployments for every pull request, providing a unique MCP URL for reviewers to test changes with real clients before merging.
Key capabilities include extensive publishing checks that align with marketplace requirements across six categories: protocol and discovery, tool conformance, security and policy, metadata and configuration, domain/TLS/CSP, and assets. Each failed check comes with fix guidance and an autofix flow. End-to-end checks run the server live within ChatGPT and Claude to verify tool calls and widget rendering, and a generated submission pack produces listing copy, tool justifications, and reviewer test cases.
Manufact's evaluation (evals) system allows developers to create test suites for their MCP servers. These evals run tasks in real clients and coding-agent harnesses across chosen models (GPT, Claude, Gemini) and clients (ChatGPT, Claude, Inspector, Claude Code, Codex CLI, Pi). An LLM judge scores each run based on recorded evidence, including tool calls, arguments, agent replies, and rendered MCP App screenshots. Evals can be integrated into CI/CD pipelines to gate merges and can be run on demand before releases, providing reports and fix prompts for regressions.
For ongoing operations, Manufact offers detailed MCP-first analytics, tracking tool calls, traces, error rates with issue triage, and latency metrics (p50/p95/p99) per tool, resource, and prompt. Analytics are broken down by client (ChatGPT, Claude, Cursor), version, and country. The platform also reveals real user goals and feedback sent through a `send_feedback` tool. Manufact supports connecting existing remote MCP servers on its Enterprise plan, providing evals, publishing checks, and submission packs without requiring hosting on Manufact. The platform also maintains the open-source `mcp-use` framework, which aids in building MCP servers and apps.
- CategoryAI DevOps & Cloud Tools
- Pricing ModelFreemium
- DeploymentCloud
- PlatformsWeb
Manufact's Core Features
Deploy MCP servers from GitHub or CLI
Run publishing checks for plugin marketplaces
Generate submission packs for ChatGPT, Claude, and other marketplaces
Monitor usage with MCP-first analytics
Run evals against various harnesses, models, and clients
Support for any language and MCP framework
Preview deployments for pull requests
Six categories of publishing checks with autofix flows
End-to-end checks within ChatGPT and Claude
LLM judge for eval scoring based on evidence
Integration of evals into CI/CD pipelines
Detailed analytics on tool calls, errors, and latency
AI-discovered user goals and feedback tracking
Support for external MCP server connections (Enterprise plan)
Open-source `mcp-use` framework for building MCP servers
How to use Manufact?
Deploy Server: Connect GitHub App for automatic deployments or use `npx mcp-use deploy` from the CLI to get a live MCP endpoint.
Run Publishing Checks: Utilize Manufact's six categories of checks to ensure compliance with marketplace requirements, receiving fix guidance for any failures.
Generate Submission Pack: Produce a complete submission pack, including listing copy, tool justifications, and reviewer test cases, for marketplaces like ChatGPT and Claude.
Configure Evals: Define test suites with representative tasks, rubrics, and passing thresholds to evaluate MCP server performance across various clients and models.
Monitor Analytics: Access detailed MCP-first analytics to track tool calls, error rates, latency, and user goals, broken down by client, version, and country.
Integrate into CI/CD: Gate continuous integration with eval suites to prevent regressions and run on-demand tests before releases.
Manufact's Use Cases
- Deploying MCP Servers
- Marketplace Publishing
- Plugin Quality Assurance
- Continuous Integration Testing
- Performance Monitoring
- User Behavior Analysis
- Cross-Client Compatibility
- Developer Workflow Optimization
- Enterprise Plugin Management







