AgentBase MCP Server
# AgentBase MCP Server
[](https://github.com/punkpeye/awesome-mcp-servers)
[](https://glama.ai/mcp/servers/AgentBase1/mcp-server)
MCP server for the [AgentBase](https://agentbase-registry.vercel.app) agent instruction registry. Gives any MCP-compatible AI agent direct access to search and retrieve instruction files.
**Listed in [punkpeye/awesome-mcp-servers](https://github.com/punkpeye/awesome-mcp-servers) — 83k+ stars.**
## What It Does
AgentBase is an open registry of agent instruction files: system prompts, skills, workflows, domain packs, safety filters, and orchestration patterns. This MCP server wraps the registry's HTTP endpoints and exposes them as structured MCP tools.
## Tools
| Tool | Description |
|------|-------------|
| `search_registry` | Search instruction files by keyword, domain, or type |
| `get_file` | Retrieve a specific instruction file by ID |
| `list_categories` | List all available categories and domains |
| `get_featured` | Get featured/top-rated instruction files |
## Install
```json
{
"mcpServers": {
"agentbase": {
"command": "npx",
"args": ["-y", "@agentbase1/mcp-server"]
}
}
}
```
## Use Cases
- AI agents that need to load domain-specific instructions at runtime
- Claude Code workflows that pull skills from a shared registry
- Teams sharing system prompts and workflows across agents
- Safety filter retrieval for production AI deployments
## Registry
Browse the full registry at [agentbase-registry.vercel.app](https://agentbase-registry.vercel.app).
CC0 licensed. All instruction files are free to use without restriction.
## Built By
[Chris Izworski](https://chrisizworski.com) — Solutions Consultant at [Prepared](https://www.prepared911.com), former Michigan 911 Executive Director, and AI implementer. Bay City, Michigan.
- Website: [chrisizworski.com](https://chrisizworski.com)
- GitHub: [github.com/izworskic](https://github.com/izworskic)
- Wikidata: [Q138283432](https://www.wikidata.org/wiki/Q138283432)
TDQS
Scored across 4 tools
Each tool has a clearly distinct purpose with no overlap: get_featured retrieves high-quality files, get_instruction fetches a specific file by slug, list_categories shows available categories, and search_registry performs keyword searches. The descriptions reinforce these unique roles, making tool selection straightforward for an agent.
All tool names follow a consistent verb_noun pattern (get_featured, get_instruction, list_categories, search_registry) using snake_case throughout. This predictability enhances readability and usability, with no deviations in style or convention.
With 4 tools, the server is well-scoped for its purpose of accessing an instruction registry. Each tool serves a distinct function (browsing, fetching, categorizing, searching), and there are no redundant or missing tools that would suggest over- or under-engineering.
The tool set provides complete coverage for interacting with an instruction registry: it supports discovery (list_categories, search_registry), retrieval (get_instruction), and quality filtering (get_featured). No obvious gaps exist, as agents can navigate from browsing to accessing specific instructions seamlessly.