toolsdk-mcp-registry

An open-source “MCP Registry & Gateway” that lets clients discover and execute Model Context Protocol (MCP) servers through a unified HTTP API. It supports bridging STDIO-based local MCP servers and remote Streamable HTTP servers, includes a sandboxed execution model, and provides an OAuth 2.1 proxy. It also exposes a publicly hosted registry dataset/index via JSON for programmatic discovery (data-only usage).

Evaluated Mar 30, 2026 (22d ago)
Homepage ↗ Repo ↗ Infrastructure mcp mcp-registry gateway oauth sandbox streamable-http stdio-bridge meilisearch docker toolsdk ai-agent
⚙ Agent Friendliness
59
/ 100
Can an agent use this?
🔒 Security
57
/ 100
Is it safe for agents?
⚡ Reliability
30
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
78
Documentation
70
Error Messages
0
Auth Simplicity
65
Rate Limits
25

🔒 Security

TLS Enforcement
80
Auth Strength
70
Scope Granularity
30
Dep. Hygiene
45
Secret Handling
55

Provides sandboxed execution and an OAuth 2.1 proxy, suggesting attention to isolating third-party tools and supporting authenticated flows. However, the provided content does not specify TLS requirements, auth scope granularity, secure secret-handling practices (e.g., logging/redaction), or detailed security headers/rate limiting behavior.

⚡ Reliability

Uptime/SLA
0
Version Stability
55
Breaking Changes
30
Error Recovery
35
AF Security Reliability

Best When

You want a centralized, self-hosted gateway/registry to execute many MCP tools securely (with sandboxing) and to integrate discovery/execution into agent systems via HTTP.

Avoid When

You cannot operate a Dockerized service (and related dependencies like search) or you need a lightweight local-only MCP client with no registry/gateway overhead.

Use Cases

  • Self-host an MCP tool gateway for LLM/agent applications
  • Remote execution of MCP tools via a single REST-style HTTP endpoint
  • Expose STDIO-only MCP servers to clients that only speak Streamable HTTP
  • Federated search/discovery across registered MCP packages/servers
  • Sandboxed execution of untrusted third-party MCP servers in production

Not For

  • Direct ad-hoc tool execution without running/staging the gateway
  • Environments that require strong guarantees about third-party tool safety beyond sandboxing
  • Use cases that need a pure MCP-native client with no HTTP translation layer

Interface

REST API
Yes
GraphQL
No
gRPC
No
MCP Server
No
SDK
Yes
Webhooks
No

Authentication

Methods: OAuth 2.1 proxy (for agent integration flows) Per-request OAuth integration via the gateway (as described in docs references)
OAuth: Yes Scopes: No

README indicates built-in OAuth 2.1 support and a proxy. Specific auth headers/parameter names and granular scopes are not shown in the provided content.

Pricing

Free tier: No
Requires CC: No

The repo/package appears open-source and self-hosted; no pricing tiers are stated in the provided README content.

Agent Metadata

Pagination
none
Idempotent
False
Retry Guidance
Not documented

Known Gotchas

  • Tool execution is mediated by the gateway; session-based MCP over Streamable HTTP requires reusing the provided mcp-session-id header (expires after ~30 minutes).
  • Environment-variable passthrough for MCP tools uses x-mcp-env-* headers; agents must avoid leaking secrets into logs.
  • Sandbox can be disabled via configuration (MCP_SANDBOX_PROVIDER=LOCAL) which may be unsafe for untrusted tools.

Alternatives

Full Evaluation Report

Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for toolsdk-mcp-registry.

AI-powered analysis · PDF + markdown · Delivered within 30 minutes

$99

Package Brief

Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.

Delivered within 10 minutes

$3

Score Monitoring

Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.

Continuous monitoring

$3/mo

Scores are editorial opinions as of 2026-03-30.

8642
Packages Evaluated
17761
Need Evaluation
586
Need Re-evaluation
Community Powered