Manus Mcp Gateway
БесплатноНе проверенProduction Pure Go Model Context Protocol (MCP) server for Manus AI v2 with Credit-Aware Capacity Pool
Описание
Production Pure Go Model Context Protocol (MCP) server for Manus AI v2 with Credit-Aware Capacity Pool
README
CI Go Report Card License: MIT PRs Welcome Latest Release
Production-grade, zero-dependency Pure Go Model Context Protocol (MCP) server for Manus AI (api.manus.ai v2).
Features a Credit-Aware Capacity Pool across multiple API keys, greedy load balancing, failover handling, and automatic credit protection for autonomous AI swarms.
⚡ Highlights
- Pure Go & Zero Dependencies: Compiles to a single lightweight binary (~8MB) with zero external C/runtime dependencies.
- Multi-Key Capacity Pool: Aggregates arbitrary numbers of Manus API accounts (e.g. 7+ accounts, 2100+ daily refresh credits).
- Greedy Credit Routing: Automatically dispatches new tasks to the account with the highest spendable credits.
- Transparent Rate-Limit Failover: Gracefully switches to alternative keys upon HTTP 429 with
Retry-Afterjitter backoff. - Credit Protection Kill-Switch: Instant task cancellation via
manus_stop_taskto prevent runaway credit drain. - Native MCP Stdio: Plugs seamlessly into Claude Desktop, Cursor, Antigravity, and
mcp-router.
🏗️ Architecture
┌────────────────────────────────────────┐
│ Autonomous AI Agents / Clients │
│ (Claude, Cursor, Antigravity, Swarms) │
└───────────────────┬────────────────────┘
│ JSON-RPC (stdio)
▼
┌────────────────────────────────────────┐
│ manus-mcp-gateway │
│ │
│ ┌──────────────────────────────────┐ │
│ │ FastMCP Tool Handlers │ │
│ └──────────────────┬───────────────┘ │
│ │ │
│ ┌──────────────────▼───────────────┐ │
│ │ Credit-Aware Capacity Pool │ │
│ │ (Greedy Routing & Failover) │ │
│ └──────────────────┬───────────────┘ │
│ │ │
│ ┌──────────────────▼───────────────┐ │
│ │ Manus API v2 Client │ │
│ └──────────────────┬───────────────┘ │
└─────────────────────┼──────────────────┘
│ HTTPS / REST (Keep-Alive)
▼
┌──────────────────────┐
│ api.manus.ai │
│ (Manus Cloud VM) │
└──────────────────────┘
🧰 Tool Suite & Operational Tiers
manus-mcp-gateway provides a complete toolset for delegating and monitoring cloud workloads. To prevent context window pollution and tool selection ambiguity, tools are categorized into three operational tiers:
| Tool Name | Operational Tier | Description | Key Parameters |
|---|---|---|---|
manus_create_task |
🟢 Tier 1: Recommended Core | Launches an asynchronous autonomous task on Manus cloud VM with greedy capacity routing. | prompt (req), title, agent_profile, key_id |
manus_get_task_status |
🟢 Tier 1: Recommended Core | Polls progress, retrieves assistant commentary and generated artifact download links. | task_id (req), order, limit, key_id |
manus_send_message |
🟡 Tier 2: Interactive Session | Sends follow-up instructions or user answers to an active session. | task_id (req), content (req), key_id |
manus_stop_task |
🟡 Tier 2: Safety Kill-Switch | Aborts a running task immediately to halt credit consumption. | task_id (req), key_id |
manus_get_pool_status |
🔴 Tier 3: SRE / Admin | Real-time breakdown of all keys, available credits, refresh dates, and pool total. | (none) |
🎯 Recommended Agent Configuration (Occam's Razor)
For 90% of autonomous coding agents, Telegram bots, and assistants (Claude, Cursor, Antigravity Swarm), exposing all 5 tools is unnecessary and counterproductive.
Applying Occam's Razor: the agent only needs to delegate work and retrieve the final artifact.
Recommended Minimal Profile (Tier 1: 2 Tools)
manus_create_taskmanus_get_task_status
Why?
- Context Window Efficiency: Saves thousands of tokens per turn on unused schemas.
- Deterministic Routing: Eliminates agent confusion (e.g. an agent trying to call
manus_get_pool_statusinstead of fulfilling a user prompt). - Fire & Check Pattern: The agent fires the task, reports the task ID, and checks completion upon user request.
🗺️ Complete Verified API Map & Tool Hygiene Guidelines
Curious about the entire Manus API v2 ecosystem? We conducted a rigorous empirical live probe of all 31 known endpoints against api.manus.ai to verify their existence, response formats, and argument constraints:
⚠️ Why Exposing All 31 Endpoints as MCP Tools is Harmful
- Token Tax & Context Drain: 30+ tool schemas inject 4,000–8,000 tokens of static descriptions on every single LLM turn.
- Tool Selection Hallucinations: High decision entropy causes models to invoke administrative CRUD tools (
agent.update,project.create,browser.onlineList) instead of executing user tasks. - Administrative Traps: Agents get trapped in recursive discovery and inspection loops.
- Destructive Blast Radius: Exposing endpoints like
task.delete,file.delete, andwebhook.deleteto autonomous agents invites accidental data loss.
⭐ Curated Must-Have Toolset
- Tier 1 (Core Must-Have — 95% of use cases):
manus_create_task,manus_get_task_status. - Tier 2 (Interactive Workflows & Pipelines):
manus_send_message,manus_confirm_action(roadmap),manus_stop_task,manus_upload_file. - Tier 3 (Cluster Infrastructure SRE):
manus_get_pool_status. - Excluded (Human / REST-only):
agent.*,project.*,skill.*,browser.*,webhook.*,website.*,*.delete.
🚀 Quickstart & Installation
Option A: Install via Go
go install github.com/TheNovaNodes/manus-mcp-gateway/cmd/manus-mcp-gateway@latest
Option B: Pre-built Binary
Download the latest pre-compiled binary for Linux, macOS, or Windows from Releases.
Option C: Build from Source
git clone https://github.com/TheNovaNodes/manus-mcp-gateway.git
cd manus-mcp-gateway
make build
# Binary is ready at bin/manus-mcp-gateway
Option D: Docker
docker build -t manus-mcp-gateway .
docker run -i --rm -e MANUS_KEYS="acc1 sk-..." manus-mcp-gateway
⚙️ Configuration & Integrations
1. Configure Keys
The gateway supports single-key or multi-account modes via environment variables:
# Multi-Account Pool (Email + Key pairs for distinct tenants):
export MANUS_KEYS="[email protected] sk-key1 [email protected] sk-key2"
# Or comma-separated keys:
export MANUS_KEYS="sk-key1,sk-key2"
# Single-key mode:
export MANUS_API_KEY="sk-your-single-key"
[!NOTE] Principal-Aware Routing vs Shared Limits: Manus enforces rate limits per user/account. Multiple keys belonging to the same account share a single rate-limit bucket. True additive throughput and capacity pooling is achieved when configuring keys across distinct, authorized accounts (
[email protected] sk-... [email protected] sk-...).
You can also store keys in a local .env file (chmod 600 .env):
cp .env.example .env
2. Claude Desktop Integration
Add to claude_desktop_config.json:
{
"mcpServers": {
"manus": {
"command": "/path/to/manus-mcp-gateway",
"env": {
"MANUS_KEYS": "[email protected] sk-key1 [email protected] sk-key2"
}
}
}
}
3. mcp-router Integration
Add to your mcp-router/config.yaml:
servers:
manus-gateway:
transport: stdio
command: /usr/local/bin/manus-mcp-gateway
args: []
env:
MANUS_KEYS: ${MANUS_KEYS}
prefix: manus__
agent_tools_allowlist:
worker_agent:
- "manus_create_task"
- "manus_get_task_status"
supervisor_agent:
- "manus_create_task"
- "manus_get_task_status"
- "manus_send_message"
- "manus_stop_task"
- "manus_get_pool_status"
🔬 Connectors (Research & Exploratory Phase)
Manus supports external connectors (both Custom MCP servers and built-in SaaS integrations like Instagram, Slack, etc.). Within NovaNodes, connectors are currently classified as Exploratory / Research features:
Custom MCP Connectors (Bi-Directional Cognitive Bridge):
- Enables cloud Manus agents to reach back into local stacks (e.g.
Nextcloud,AnythingLLM, private Git) over Streamable HTTP. - Requires API v2
message.connectorsmapping viaGET /v2/connector.listIDs. - Status: Research prototype in progress (tracked under Draft PR #5 & Issue #4).
- Enables cloud Manus agents to reach back into local stacks (e.g.
Instagram Connector (
4b899211-fd12-410e-a8d2-264a409cbc78):- Native Meta Content Publishing API integration (Posts, Carousels, Reels, Stories, Insights).
- Requires Instagram Professional (Creator/Business) account linked to a verified Facebook Page with Admin Full Control.
- Status: Backlogged pending Meta OAuth & network stabilization (tracked under Issue #6).
🛡️ Quality Gate & Testing
# Run unit tests with race detection
make test-race
# Run static analysis
make lint
# Clean build artifacts
make clean
🤝 Contributing
Contributions, issues, and feature requests are welcome!
Please see our CONTRIBUTING.md guide and Code of Conduct.
For security reports, review SECURITY.md.
📄 License
This project is licensed under the MIT License.
Установка Manus Mcp Gateway
У этого сервера нет опубликованного пакета — он собирается из исходников. Открой репозиторий и следуй инструкции в README.
▸ github.com/TheNovaNodes/manus-mcp-gatewayFAQ
Manus Mcp Gateway MCP бесплатный?
Да, Manus Mcp Gateway MCP бесплатный — установка в пару кликов через Unyly без оплаты.
Нужен ли API-ключ для Manus Mcp Gateway?
Нет, Manus Mcp Gateway работает без API-ключей и переменных окружения.
Manus Mcp Gateway — hosted или self-hosted?
Self-hosted: сервер запускается локально на твоей машине командой из раздела установки.
Как установить Manus Mcp Gateway в Claude Desktop, Claude Code или Cursor?
Открой Manus Mcp Gateway на unyly.org, выбери вкладку своего клиента (Claude Desktop, Claude Code, Cursor) и нажми Install — конфиг сгенерируется автоматически, без правки JSON.
Похожие MCP
Fetch
Web content fetching and conversion for efficient LLM usage.
Roblox Studio
Enables AI coding tools to control Roblox Studio for workspace exploration, instance manipulation, and script management. It provides tools for playtesting, sce
автор: paralovOpencode Omniroute Plugin
OpenCode plugin for the OmniRoute AI Gateway. Drives dynamic model discovery, /connect auth flow, and multi-instance OmniRoute providers via the official @openc
автор: GitHub ActionsAWS KB Retrieval
Retrieval from AWS Knowledge Base using Bedrock Agent Runtime.
автор: modelcontextprotocolSpring AI MCP Server
Provides auto-configuration for setting up an MCP server in Spring Boot applications.
llm-analysis-assistant
A very streamlined mcp client that supports calling and monitoring stdio/sse/streamableHttp, and can also view request responses through the /logs page. It also
автор: xuzexin-hzMCP-Agent
A simple, composable framework to build agents using Model Context Protocol by [LastMile AI](https://www.lastmileai.dev)
автор: lastmile-aiSpring AI MCP Client
Provides auto-configuration for MCP client functionality in Spring Boot applications.
mcp.natoma.ai
A Hosted MCP Platform to discover, install, manage and deploy MCP servers by [Natoma Labs](https://www.natoma.ai)
MCPHub
Website to list high quality MCP servers and reviews by real users. Also provide online chatbot for popular LLM models with MCP server support.
Compare Manus Mcp Gateway with
Не уверен что выбрать?
Найди свой стек за 60 секунд
Автор?
Embed-бейдж для README
Похожее
Все в категории ai
