agentleFS
jahfaliabdulrahman-devjahfaliabdulrahman-dev/hermes-skills/skills/mcp-bridge-patterns/SKILL.md

mcp-bridge-patterns

Skill1 starsChanged 42 days agoApache-2.0View on GitHub

Zero-dependency MCP stdio server patterns — non-blocking dispatch, manifest desync fixes, shell-vs-LLM verification. For building MCP bridges that connect external LLM clients (Claude Desktop, Kimi) to Hermes Agent.

  • Reads credentials

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 0 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Retry: Only socket hang up, ECONNRESET, timeout.

    line 50
  2. Very specific:

    API key source: Read from the Hermes environment file at server startup.

    line 52
  3. Very specific:

    All tools (quick_task, lead_delegate, kanban views) should be non-blocking.

    line 58

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

MCP Bridge Patterns

Recurring patterns for building MCP servers that bridge Claude Desktop to Hermes Agent. v2.0: HTTP Bridge replaces subprocess spawning.

When to Use

  • Building a new MCP bridge (Node.js, stdio transport)
  • Debugging "MCP error -32001" or socket hang up from Hermes Gateway API
  • Adding tools to an existing MCP server
  • Routing tasks to Kanban boards based on project path

Pattern 1: HTTP Bridge (v2.0)

Problem: spawn("hermes chat -q") is slow, fragile, and fails in Claude Desktop because MCP stdio servers inherit a limited subset of PATH. hermes binary is not available.

Solution: HTTP POST to Hermes Gateway API. Zero subprocess. Retry logic handles transient failures.

function hermesApi(messages, profile, timeoutMs, retries = 2) {
  return new Promise((resolve, reject) => {
    const attempt = (n) => {
      const req = request(`${API_URL}/v1/chat/completions`, {
        method: "POST",
        headers: { "Authorization": `Bearer ${API_KEY}`, "Content-Type": "application/json" },
        timeout: timeoutMs,
      }, (res) => { /* accumulate data, parse, resolve */ });
      req.on("error", (e) => {
        if (n > 0 && isTransient(e)) setTimeout(() => attempt(n-1), backoff(n));
        else reject(e);
      });
      req.write(body); req.end();
    };
    attempt(retries);
  });
}

Retry: Only socket hang up, ECONNRESET, timeout. Backoff 2s→4s. Max 3 attempts.

API key source: Read from the Hermes environment file at server startup. Set the Authorization header on every request.


Pattern 2: Non-blocking Dispatch with File Persistence

All tools (quick_task, lead_delegate, kanban views) should be non-blocking. Return task_id immediately. Background HTTP requests persist state to disk. Client polls via lead_task_status.

Key features: file persistence (survives restarts), content-hash dedup (duplicate EPICs get same ID), max 2 concurrent, auto-cleanup after 1h.


Pattern 3: Kanban-Aware lead_task_status

lead_task_status accepts two ID formats:

IDSourceResolution
task-xxxbridge dispatchFile persistence
t_xxxKanban cardHTTP API to Lead Architect profile — NEVER exec()

Never use exec() to query Kanban. MCP servers don't inherit the user's PATH. Always route through the Gateway API with the Lead Architect profile.


Pattern 4: Board Routing via project_path

Map project_path to board slug. Inject into Lead Architect prompt:

const boardMap = { "Azdal": "azdal", "CarSah": "carsah" }; // key = project folder basename
const board = boardMap[projectPath] || "default";

Boards created with hermes kanban boards create <slug>. Isolated per-project history.


Pattern 5: project_path Enforcement

lead_delegate MUST require project_path. Validate with fs.existsSync(). No path → hard refusal. Prevents the #1 contamination class: EPICs routed to wrong project via stale memory.


Pattern 6: MCP PATH Limitations

Per MCP spec: stdio servers inherit a limited subset of environment variables. Never use exec(), spawn(), or any subprocess that depends on the user's PATH. Route everything through the Gateway API — it runs in the full user environment.


Zero-Dependency MCP Server Boilerplate (Node.js)

#!/usr/bin/env node
import { spawn, exec } from "child_process";

// Stdio JSON-RPC transport
process.stdin.on("data", data => {
  for (const line of data.toString().trim().split("\n")) {
    if (!line) continue;
    try { handleMessage(JSON.parse(line)); } catch(e) { /* log to stderr */ }
  }
});

function send(msg) { process.stdout.write(JSON.stringify(msg) + "\n"); }

function handleMessage(msg) {
  if (msg.method === "initialize") {
    send({ jsonrpc: "2.0", id: msg.id, result: {
      protocolVersion: "2024-11-05",
      capabilities: { tools: {} },
      serverInfo: { name: "...", version: "1.0.0" }
    }});
  } else if (msg.method === "tools/list") {
    send({ jsonrpc: "2.0", id: msg.id, result: { tools: [...] }});
  } else if (msg.method === "tools/call") {
    runTool(msg.params.name, msg.params.arguments)
      .then(r => send({ jsonrpc: "2.0", id: msg.id, result: r }))
      .catch(e => send({ jsonrpc: "2.0", id: msg.id, result: { content: [{ type: "text", text: e.message }], isError: true }}));
  }
}

// Keep process alive (MCP stdio server must never exit)
setTimeout(() => {}, 86400000);

package.json requirements:

{ "type": "module", "main": "server/index.js" }

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.