Thanks for contributing! This guide covers how to add new MCP servers to the registry.
Copy the template and create servers/{server-id}.yaml where {server-id} matches the id field inside the file.
cp servers/_template.yaml servers/my-server.yamlNaming conventions:
- Use kebab-case: lowercase alphanumeric with hyphens
- Pattern:
^[a-z0-9]+(-[a-z0-9]+)*$ - Maximum 128 characters
- Be descriptive but concise (e.g.,
github,brave-search, notghorbs) - Filename must match the
idfield exactly
Every MCP server definition must include:
| Field | Description |
|---|---|
id |
Unique kebab-case identifier (must match filename) |
name |
Human-readable display name (max 255 chars) |
description |
What the server does and when to use it (10-2000 chars) |
author |
Object with name (required) and github (required) |
repository |
URL to the source code repository |
transport |
Array of supported protocols: stdio, sse, http-streaming |
category |
One of the 11 categories (see below) |
tags |
1-20 searchable tags in kebab-case (max 64 chars each) |
created_at |
ISO 8601 timestamp (e.g., "2025-06-01T00:00:00.000Z") |
updated_at |
ISO 8601 timestamp |
These are optional but highly recommended — they're what makes your server discoverable:
Tools (max 200 per server):
tools:
- name: tool_name # Max 128 chars
description: What it does # Max 1000 chars
input_schema:
type: object
properties:
param:
type: string
description: Parameter description
required:
- paramResources (max 100 per server):
resources:
- uri_template: "file:///{path}" # Max 2048 chars
description: What it provides # Max 1000 charsPrompts (max 100 per server):
prompts:
- name: prompt_name # Max 128 chars
description: What it does # Max 1000 chars
arguments:
- name: arg_name # Max 128 chars
description: Arg desc # Max 500 chars| Field | Description |
|---|---|
install |
Installation config (npm, command, args) |
env_vars |
Environment variables needed (name, description, required) |
license |
SPDX license identifier (e.g., MIT, Apache-2.0) |
version |
Current version of the MCP server |
stars |
GitHub star count (updated periodically by maintainers) |
pip install -r requirements.txt
python validate.pyValidation checks:
- YAML syntax is valid
- All required fields are present and correctly typed
- Schema constraints are met (string lengths, patterns, enums)
- Filename matches the
idfield - No duplicate IDs across all server definitions
- Fork this repository
- Create your server YAML file in
servers/ - Validate locally with
python validate.py - Commit with a descriptive message (e.g.,
Add my-server MCP server definition) - Open a PR with a brief description of the server's purpose
Fork repo → Create YAML → Validate locally → Open PR → CI validates → Review → Merge
- Automated validation — GitHub Actions runs
validate.pyon your PR. If validation fails, a comment is posted with the specific errors. - Review — A maintainer reviews the submission for quality and accuracy.
- Merge — Once validation passes and a maintainer approves, the PR is merged.
- Ingestion — On the next deployment, your server is ingested into the platform and appears on openmodels.run/mcp.
Maintainers check for:
- Accuracy — does the server actually exist and work as described?
- Completeness — are tools, resources, and prompts documented?
- Quality — are descriptions clear and helpful?
- No duplicates — is this server already in the registry under a different ID?
- Valid repository — does the repository URL point to a real, accessible project?
Choose the most appropriate category for your server:
| Category | Description |
|---|---|
| Development | Code generation, testing, debugging, and software development tools |
| Productivity | Task management, note-taking, workflow automation |
| Database | Database querying, schema management, and data manipulation |
| DevOps | CI/CD, infrastructure management, monitoring |
| AI | Machine learning, model management, and AI utilities |
| Research | Web search, knowledge retrieval, and information gathering |
| Browser Automation | Web scraping, browser control, and automated testing |
| Communication | Messaging, email, notifications, and collaboration |
| Filesystem | File operations, directory management, and local storage |
| Cloud | Cloud provider APIs, serverless functions, and cloud storage |
| Security | Authentication, encryption, and vulnerability scanning |
See the full schema at schemas/server.schema.json and the annotated template at servers/_template.yaml.
Key constraints:
id: kebab-case, max 128 chars, must match filenamename: 1-255 charsdescription: 10-2000 charstransport: array of 1-3 values fromstdio,sse,http-streamingtags: 1-20 kebab-case strings, each max 64 charstools: max 200 entriesresources: max 100 entriesprompts: max 100 entries- Timestamps: ISO 8601 format (
YYYY-MM-DDTHH:MM:SS.sssZ)
- Be specific in descriptions — explain what problems the server solves
- Document all tools — users discover servers by their tool capabilities
- Include install config — makes it easy for users to get started
- List environment variables — users need to know what credentials are required
- Use accurate tags — they power search and discovery
- One server per PR — easier to review and faster to merge
Be respectful, constructive, and inclusive. We're building a community resource.