Audit report
anysearch-ai/anysearch-mcp-server
https://github.com/anysearch-ai/anysearch-mcp-server
Type: mcp · Audited 2026-05-22
Top fixes
- high
Add a server.json manifest
Add a server.json manifest. It is the machine-readable declaration of your server that the MCP registry and agents read directly.
- high
Expand the Tools section
Expand the Tools section. Each tool needs a description of what it does and when an agent should call it, not just a name.
- high
Add usage examples showing a real tool call
Add usage examples showing a real tool call. Examples are how an agent learns the expected inputs.
- high
Add an open-source license
Add an open-source license. Some agents will not run or recommend code with no SPDX license.
- high
Get the server into public MCP config files
Get the server into public MCP config files. Few or none reference it today, and real adoption is the strongest trust signal an agent has.
Sub-scores
Discoverability
Agents can only recommend tools they can find. llms.txt and open robots.txt are the fastest wins.
Passing
- ✓Repository has a clear description (52)
- ✓Repository has topics/tags (6)
- ✓Identifiable as an MCP server
- ✓README opens with a one-line summary
Installation
Config-file appearances are the closest thing to a real install count for MCP servers.
Missing
- ✗Ships a package manifest
Passing
- ✓README shows an MCP client config block
- ✓README has an install or run command
- ✓Transport or configuration is documented
Tool definitions
Structured markup lets agents extract facts reliably instead of guessing from prose.
Missing
- ✗Ships a server.json manifest
- ✗Tools are described, not just named
- ✗README shows usage examples
Passing
- ✓README enumerates the tools
Documentation
Thin or poorly-described content gives agents nothing to summarise or cite.
Missing
- ✗README states what the server is for
- ✗Versioned releases or a changelog
Passing
- ✓Substantive README (1200+ chars) (9938)
- ✓README is organized into sections (28)
Trust & adoption
Models pattern-match provenance signals before recommending — missing basics trigger silent skips.
Missing
- ✗Open-source license (none)
- ✗Appears in public MCP config files (0)
- ✗Has tests or CI
- ✗Has tagged releases (none)
- ✗Discussed on Hacker News or Reddit (0)
Passing
- ✓Pushed in the last 90 days (0)
- ✓Multiple contributors (2)
- ✓25+ GitHub stars (460)
Generated llms.txt
Drop this at /llms.txt on your domain.
Embed badge
Paste this into your README to display your live score.
[](https://mcprank-green.vercel.app/r/github.com-anysearch-ai-anysearch-mcp-server-9b40ae)
Get notified when this changes
We re-audit periodically. We'll email you when this score moves.
Discovery Report · free
The audit tells you whether agents CAN pick this server. Want to know whether they actually DO?
The Discovery Report measures which queries agents route to this server vs its rivals across multiple AI model families. It picks the one intent where you have the most leverage and gives you a concrete change to ship this week. Free.