electrical_services

Aerostack MCP Server — Hosted Developer Tools Integration

MCP Server language Hosted language Public

Build, deploy, and orchestrate AI integrations — plan connections, auto-generate workflows, create private MCPs, and publish to marketplace. 17 tools for end-to-end automation.

@aerostack verified

• v0.1.0 • MIT • Updated May 13, 2026

add_circleAdd to Workspace

robot_2

Use with AI AssistantsMCP

Connect Claude, Cursor, or any MCP-compatible client — then call tools directly

① Add This MCP Server

Paste into your AI client config — then all its tools are available instantly.

.claude/mcp.json

{
  "mcpServers": {
    "aerostack": {
      "url": "https://mcp.aerostack.dev/s/aerostack/mcp-aerostack",
      "headers": {
        "Authorization": "Bearer YOUR_AEROSTACK_TOKEN"
      }
    }
  }
}

Replace YOUR_AEROSTACK_TOKEN with your API token from the dashboard.

② Call a Tool

Ask your AI assistant to call a specific tool, or send raw JSON-RPC:

+16 more

Natural Language Prompt

“Use the guide tool to get help on aerostack. topics: start, credentials, telegram_setup, discord_setup, whatsapp_setup, slack_setup, llm_keys, workspace_tools, functions, integrations, node:<type>. node types: trigger, llm_call, logic, mcp_tool, send_message, action, loop, code_block, auth_gate, schedule_message, delegate_to_bot, send_proactive, error_handler, parallel. use guide("start") first if new to aerostack”

Using a Workspace?

Add this MCP to your Workspace — your team shares one token, secrets are stored securely, and every AI agent in the workspace can call it without per-user setup.

add_circleAdd to Workspace

description Overview

mcp-aerostack — Aerostack Developer MCP

Turn any AI coding agent into a full-stack Aerostack developer. Plan integrations, build bots, generate workflows, create private MCPs, and publish to marketplace — all through conversation.

Aerostack is a developer platform for building AI-native backends on the edge. This MCP server gives your AI agent (Claude Code, Cursor, Windsurf, or any MCP-compatible tool) the ability to plan and execute end-to-end integrations — from analyzing what services are needed, through building missing connectors, to deploying production workflows.

Instead of clicking through a dashboard or writing boilerplate, just describe what you want: "Connect my Notion to Ocoya — post new pages as social media content every hour" — and your AI agent handles everything.

What You Can Do

Orchestrate Integrations (NEW)

Plan — Describe an integration in plain English. The agent searches the marketplace, detects gaps (missing MCPs), determines credential requirements, and returns a structured execution plan.
Build Missing MCPs — When a service isn't in the marketplace, get a complete MCP template. Your AI editor writes the implementation, deploys as a private MCP.
One-Call Setup — Create workspace, add MCPs, store encrypted secrets, verify health — all in one step.
Auto-Generate Workflows — Agent discovers available tools, builds the trigger → read → transform → write pipeline automatically.
Publish to Marketplace — Share your private MCPs with the community. Quality standards enforced automatically.

Build & Deploy

AI Bots — Telegram, Discord, WhatsApp, and Slack bots with LLM-powered workflows.
Visual Workflows — 16 node types: LLM calls, logic, loops, MCP tools, code blocks, auth gates, scheduled messages, human handoffs.
Custom Functions — TypeScript functions deployed to the edge with full runtime access (Cache, Database, AI, Queue, Vector Search, Storage).
AI Endpoints — Expose workflows as REST APIs with manual, scheduled, and webhook triggers.

Available Tools (17)

CRUD Tools (1-10)

Tool	Description
`guide`	Contextual help — platform setup, node schemas, credential requirements, runtime APIs. 18+ topics.
`list`	List resources: bots, workflows, endpoints, webhooks, functions, workspace tools, templates.
`get`	Get full details — config, workflow graph, credential status, deployment state.
`create`	Create resources. Bot credentials validated + encrypted. Functions auto-deployed.
`update`	Partial updates. Functions auto-redeployed. Workspace secrets encrypted (AES-256-GCM).
`delete`	Delete with reference checking — won't delete a workflow used by a bot.
`validate`	Pre-flight checks: graph validation, bot config, function code safety.
`test`	Execute with test input — per-node execution logs for workflows.
`deploy`	Go live: register webhooks, publish workflows, push functions to edge.
`scaffold`	Generate complete resource from natural language description.

Orchestration Tools (11-17)

Tool	Description
`plan`	The brain — Analyze integration request, search marketplace, detect gaps, return execution plan with credentials + workflow shape.
`capabilities`	Get MCP tools grouped by read/write/search, auth requirements, config schema.
`build_spec`	Generate MCP server template for AI editors — boilerplate, tool stubs, aerostack.toml, README.
`sdk_docs`	SDK reference: MCP templates, function templates, auth patterns, deploy/test guides.
`setup_integration`	One-call workspace setup: create workspace + add MCPs + store secrets + verify health.
`generate_workflow`	Auto-build workflow from workspace tools: trigger → source → transform → destination.
`publish`	Publish private MCP to marketplace with quality gate validation.

Integration Flow

1. plan("connect Notion to Ocoya")
   → Finds Notion MCP ✓, Ocoya not found ✗
   → Returns: credentials needed, workflow shape, next steps

2. build_spec({ service: "ocoya", tools: ["create_post"] })
   → Returns MCP template for your AI editor to implement

3. create({ type: "function", config: { code: "...", visibility: "private" } })
   → Deploys private MCP to edge

4. setup_integration({ name: "Notion → Ocoya", servers: [...], secrets: {...} })
   → Creates workspace, adds MCPs, stores encrypted secrets

5. generate_workflow({ workspace_id: "...", intent: "Post Notion pages to Ocoya" })
   → Auto-builds trigger → read → transform → write pipeline

6. test + deploy → Live!

7. publish({ function_id: "..." })
   → Optional: share your Ocoya MCP on the marketplace

Configuration

Variable	Required	Description
`AEROSTACK_API_KEY`	Yes	Account key (starts with `ak_`, never expires). Get from aerostack.dev → Settings → CLI Keys.

Quick Start

Connect Your AI Agent

Claude Code (.claude/mcp-config.json):

{
  "mcpServers": {
    "aerostack": {
      "url": "https://mcp.aerostack.dev/ws/YOUR_WORKSPACE_SLUG",
      "headers": {
        "Authorization": "Bearer YOUR_WORKSPACE_TOKEN"
      }
    }
  }
}

Example Conversations

Plan an integration:

You: "I want to sync GitHub issues to Linear whenever they're labeled as 'bug'"

Agent: → plan() finds both MCPs available
       → setup_integration() creates workspace with both MCPs
       → generate_workflow() builds: webhook trigger → github.get_issue → transform → linear.create_issue
       → Asks for GitHub token + Linear API key
       → Deploys workflow

Build a missing MCP:

You: "Connect Notion to Ocoya for social media posting"

Agent: → plan() finds Notion ✓, Ocoya ✗
       → build_spec("ocoya") gives template
       → Writes Ocoya MCP implementation
       → Deploys as private MCP
       → Wires workflow
       → Suggests publishing to marketplace

Deployment

This MCP is deployed privately from the Aerostack monorepo:

# Build + deploy to dev
AEROSTACK_API_KEY=ak_... npm run deploy:dev

# Build + deploy to production
AEROSTACK_API_KEY=ak_... npm run deploy:prod

Source code is NOT in the public aerostack-market repo.

terminal Tools (22)

Available tools on this MCP server. Each tool can be called directly from any AI agent.

terminal

guide #1

Get help on Aerostack. Topics: start, credentials, telegram_setup, discord_setup, whatsapp_setup, slack_setup, llm_keys, workspace_tools, functions, integrations, node:<type>. Node types: trigger, llm_call, logic, mcp_tool, send_message, action, loop, code_block, auth_gate, schedule_message, delegate_to_bot, send_proactive, error_handler, parallel. Use guide("start") first if new to Aerostack.

terminal

list #2

List resources. Returns summary fields only (id, name, status). Use get() for full details of a specific item.

terminal

get #3

Get details of a resource. Large fields (code, nodes_json, workflow_json) are summarized by default to save tokens. Pass include to get full content of specific fields.

terminal

create #4

Create a resource. Auto-detects workspace_id if not provided. Functions auto-deploy. Returns id + next_steps (test → deploy). Use guide() to learn required fields per type. Use validate() first to catch errors before creating.

terminal

update #5

Partial update — only send changed fields. Functions auto-redeploy. For workspace_secret: id=workspace_id, config={key, value}. For workspace_server: id=workspace_id, config={server_id, enabled?, inject_secrets?}.

terminal

delete #6

Delete a resource permanently. Cannot be undone. For workspace_server: id=workspace_id, config={server_id}.

terminal

validate #7

Client-side validation before create/update. Catches structural errors (missing nodes, invalid edges, missing fields) without an API call. The server does its own validation on create — this is a fast pre-check to avoid wasted round-trips.

terminal

test #8

Test-run a resource without deploying. Workflows: returns per-node execution log. Bots: sends a test message. Functions: deploys + invokes. Endpoints: sends a test request. Use logs() after to inspect run history.

terminal

deploy #9

Manage resource lifecycle. Default action=live. Bots: live (register webhook), pause (unregister), resume (re-register). Workflows: live (activate), pause (deactivate). Functions/endpoints: live (deploy to edge).

terminal

scaffold #10

Generate a ready-to-create config from natural language. Auto-detects platform, integrations, and runtime APIs. Returns config + workflow graph + missing credentials. Next: validate() → create() → test() → deploy().

terminal

plan #11

Analyze an integration request BEFORE building. Returns a complete infrastructure blueprint: which Aerostack modules to use (Bot, Workflow, AI Endpoint, Function, MCP, Workspace), service availability, gap specs for missing MCPs, credential checklist, workflow shape, and ordered build sequence. Handles any complexity — from simple "Notion to Slack" to multi-service orchestration with bots + workflows + approval gates.

terminal

capabilities #12

Get tools and auth requirements for a specific MCP/service. With workspace_id: fetches live tools from running MCPs. Without: returns catalog info + marketplace match. Use this to discover what tools exist before generate_workflow().

terminal

build_spec #13

Generate MCP server or function boilerplate code. Returns src/index.ts, aerostack.toml, tool stubs, and auth wiring. Use when plan() identifies a missing service. After implementing: create(type="function") → test() → publish().

terminal

sdk_docs #14

Get SDK code patterns for building on Aerostack. Use when writing MCP server or function code — returns correct boilerplate, auth patterns, and deploy commands.

terminal

setup_integration #15

One-call workspace setup: creates workspace (or reuses), adds MCPs, stores encrypted secrets, runs health check. Use after plan() identifies which MCPs and credentials are needed. Next: generate_workflow() → test() → deploy().

terminal

generate_workflow #16

Generate a complete working workflow. Use discover() first to get real IDs for source_config/dest_config. With configs: builds production-ready nodes with real database IDs, profile IDs, loop over results, status updates. Without configs: builds a scaffold. Next: test() → deploy().

terminal

publish #17

Publish a function/MCP to the marketplace. Validates quality (README, _ping tool, category, tags) — returns structured blockers if rejected. After publishing, others can find and install it.

terminal

automate #18

Intelligent automation builder. Orchestrates the full lifecycle: plan → configure → deploy. Phase 1 (plan): describe what to automate → returns architecture decision + questions. Phase 2 (configure): provide answers → sets up workspace, discovers data, returns confirmation. Phase 3 (deploy): confirmed → generates workflow, creates resources, deploys. Each phase returns structured output with next_steps. Start with automate({ description: "..." }) — the system decides whether to use a workflow, endpoint, or webhook.

terminal

logs #19

Get run history and execution logs for a resource. Shows recent runs with status, duration, errors, and per-node results. Use after test() or deploy() to debug issues.

terminal

diagnose #20

Health check for bots or workspaces. Bots: checks config, credentials, webhook, AI model, recent activity. With deep=true: sends test LLM call. Workspaces: checks each server health. Use when a bot/workspace isn't working as expected.

terminal

status #21

Quick account overview: resource counts, usage, plan info. Use before deploying to check quota, or to get a snapshot of what exists. Faster than calling list() for each type.

terminal

discover #22

Interrogate workspace MCP servers to discover available data. Calls read/list tools on each service (e.g. list_databases on Notion, list_workspaces on Ocoya) and returns structured choices. Use BEFORE generate_workflow to get real IDs. If multiple options exist (databases, profiles, workspaces), return them all so the agent can ask the user which to use.

Details

upgrade Version 0.1.0

gavel License MIT

wifi Transport streamable-http

lock Access Public

category Category Developer Tools

terminal Tools 22

language Live Endpoint

https://mcp.aerostack.dev/s/aerostack/mcp-aerostack

Sub-50ms globally · Zero cold start

Publisher

@aerostack verified

Pre-built functions for the most common MCP tool patterns. Clone, extend, and deploy.

Frequently asked questions

What is the Aerostack MCP server and what can it do? +

The Aerostack MCP server is hosted on Aerostack and exposes these tools to your AI agent: `guide`, `list`, `get`, `create`, `update`. You get one hosted URL — no self-hosting — that works from Claude, Cursor, ChatGPT, Gemini, VS Code, or any MCP-compatible client, and you can share it with your team or combine it with other MCP servers in a workspace.

Is the Aerostack MCP server hosted, or do I have to run it myself? +

It's hosted on Aerostack's edge infrastructure — you don't deploy or maintain anything. Add it to a workspace and you get one authenticated URL, with secrets encrypted, that any AI agent or editor can connect to. Use it solo or share the same URL across your whole team.

Which AI agents and editors can use the Aerostack MCP server? +

Any MCP client: Claude and Claude Code, Cursor, ChatGPT, Gemini, Windsurf, Cline, VS Code, and custom agents. Because it's one hosted URL, the same Aerostack MCP server works everywhere — and you can compose it with other MCP servers, skills, and functions behind a single workspace URL.

How do I install the Aerostack MCP server in Claude Desktop? +

Add the following to your Claude Desktop config (`claude_desktop_config.json`): ```json { "mcpServers": { "@aerostack/mcp-aerostack": { "command": "npx", "args": ["-y", "@aerostack/@aerostack/mcp-aerostack"] } } } ``` Then restart Claude Desktop and the tools will appear automatically.

How do I use the Aerostack MCP server in Cursor? +

In Cursor, open **Settings → MCP** and add: ```json { "name": "@aerostack/mcp-aerostack", "command": "npx", "args": ["-y", "@aerostack/@aerostack/mcp-aerostack"] } ``` Save and reload Cursor. The MCP tools will be available in Agent mode.

Does Aerostack MCP require authentication? +

Yes. Aerostack requires authentication. Check the MCP's documentation for the required credentials.

Aerostack MCP Server — Hosted Developer Tools Integration

description Overview

mcp-aerostack — Aerostack Developer MCP

What You Can Do

Orchestrate Integrations (NEW)

Build & Deploy

Available Tools (17)

CRUD Tools (1-10)

Orchestration Tools (11-17)

Integration Flow

Configuration

Quick Start

Connect Your AI Agent

Example Conversations

Deployment

terminal Tools (22)

Details

Publisher

Tags

Frequently asked questions