Hosting

Remote hosted (Cloudflare Workers); local deployment documented

Transport

Streamable HTTP

Endpoint

https://agent-never-give-up-mcp.askman.dev/mcp

Authentication

No credential was required for the hosted tools/list request. The server requires an MCP initialize/session handshake; MCP clients should perform this automatically.

Requirements

• MCP client supporting remote Streamable HTTP
• No local runtime required for hosted endpoint

Setup steps

1. Add `https://agent-never-give-up-mcp.askman.dev/mcp` as a remote Streamable HTTP MCP server.
2. Reconnect or reload the client.
3. Verify by using `list_scenarios` to inspect available recovery scenarios.
4. Treat the response as structured guidance; the tool does not itself inspect or fix the project unless the client agent carries out the returned protocol.

Nine exact tools verified by hosted tools/list after MCP initialize/session setup. Seven core scenario tools are registered directly; list_scenarios and get_prompt provide dynamic extended scenarios.

• list_scenarios

  List all available stuck-agent scenarios with their tier (core/extended). Core scenarios are also directly available as individual tools.

  01
• bug-fix-always-failed

  Use this when your bug fix attempts keep failing, and you need to break out of the loop with a fresh debugging strategy.

  Parameters: mode, contextSummary
  02
• logic-is-too-complex

  Use this when your reasoning is getting long, tangled, or circular, and you need a simple, structured way to reset and move forward.

  Parameters: mode, contextSummary
  03
• long-goal-partially-done

  Use this when working on a multi-step task and you have only completed part of it, risking that remaining steps are forgotten.

  Parameters: mode, contextSummary
  04
• lost-main-objective

  Use this when you remember the original goal but your current actions feel disconnected from it, and you need to realign your work with the main objective.

  Parameters: mode, contextSummary
  05
• missing-requirements

  Use this when you cannot proceed because requirements are missing or unclear, and you need to identify what information is needed.

  Parameters: mode, contextSummary
  06
• scope-creep-during-task

  Use this when your changes are expanding beyond the original task scope, and you need to refocus on what was actually requested.

  Parameters: mode, contextSummary
  07
• strategy-not-working

  Use this when you have tried the same approach multiple times without success, and you need to step back and consider a fundamentally different strategy.

  Parameters: mode, contextSummary
  08
• get_prompt

  Get prompt template for any scenario (core or extended) with optional mode. Use list_scenarios to discover available scenario IDs.

  Parameters: scenario, mode, contextSummary
  09

• Extended scenarios are accessed through list_scenarios/get_prompt rather than all being direct tools.
• Repository says public/auth-less v0; verify endpoint policy before sending sensitive project content.
• This server returns thinking/recovery protocols; it is not a tool executor or guarantee of autonomous repair.

• official documentation

  The current hosted server was initialized using MCP 2025-06-18 and a subsequent tools/list response returned nine tools, matching the catalog. Seven are fixed core recovery scenarios; list_scenarios and get_prompt expose additional extended scenarios dynamically. No credentials were required for the read-only inventory request.

• official tool reference

  The current hosted server was initialized using MCP 2025-06-18 and a subsequent tools/list response returned nine tools, matching the catalog. Seven are fixed core recovery scenarios; list_scenarios and get_prompt expose additional extended scenarios dynamically. No credentials were required for the read-only inventory request.

Documentation checked 2026-10-28T13:26:33Z. Connection has not been runtime tested.