Community Server: Resolve MCP -- Structured error recovery for AI agents (20+ services)

[unitedideas]

Server Information

• Name: Resolve MCP Server
• Repository: https://github.com/unitedideas/resolve-mcp
• Language: Python
• Homepage: https://resolve.arflow.io

Description

Resolve provides structured error recovery playbooks for AI agents. When an agent encounters an API error (rate limit, auth failure, timeout, connection error, etc.), Resolve returns a resolution playbook with:

• Error category (rate_limit, auth, payment, connection, timeout, server_error, etc.)
• Recommended action (backoff_and_retry, abort, check_credentials, etc.)
• Exact backoff schedule in seconds
• Maximum retry count
• Service-specific recovery notes
• Confidence score

Supported Services (20+)

anthropic, aws, cloudflare, discord, docker, gemini, github, huggingface, linear, openai, opensrs, pinecone, postgres, redis, resend, sendgrid, shopify, stripe, supabase, twilio

Tools

Tool	Description
resolve_error	Look up a resolution for a service + error code/message
list_services	List all supported services (no auth required)
submit_feedback	Report whether a resolution helped
get_free_key	Provision a free API key (500 requests/month)

Install

pip install resolve-mcp
claude mcp add resolve -e RESOLVE_API_KEY=rsl_your_key -- python -m resolve_mcp

Why

AI agents that call external APIs need to handle errors gracefully. Most either retry blindly or give up entirely. Resolve gives agents the knowledge to recover correctly -- which errors are transient vs permanent, what the optimal backoff schedule is, and what specific recovery steps to take.

Pricing

Free tier: 500 requests/month. Pro and Fleet plans available.