Buildintegrationswithyourcodingagent
Point Claude Code, Cursor, Codex, or Copilot at Unified.to. Your agent adds 940+ integrations to your app, manages your workspace, and debugs its own work.
Paste this into your coding agent:
Add Unified.to integrations to my app: https://unified.to/SKILL.mdConnect the Unified.to Core MCP server
Give your agent direct access to your Unified.to workspace. It can check which integrations support what you need, create connections and webhooks, answer questions from the docs, and read API-call logs to fix a failing integration on its own.
- · Streamable HTTP endpoints in the US and EU
- · Authenticated with your workspace API key
- · Works on workspace configuration only — never your customers' data
claude mcp add --transport http unified-core https://mcp-api.unified.to/mcp \ --header "Authorization: bearer $UNIFIED_API_KEY"
What your agent can do
Ask in plain language. The Core MCP server turns it into workspace actions.
Search the docs
Answers API questions from docs.unified.to instead of guessing.
“How does pagination work for HRIS employees?”
Check integration support
Lists the integrations and features available to your workspace.
“Which CRMs support deal webhooks?”
Manage connections
Creates, inspects, and updates connections, and generates scoped MCP URLs.
“Show me the connections for user 42.”
Set up webhooks
Registers and updates webhook subscriptions for real-time sync.
“Send updated CRM contacts to /api/webhooks/unified.”
Debug failures
Reads API-call logs and workspace issues to find the root cause.
“Why did yesterday's Salesforce sync fail?”
Manage environments
Creates and lists environments so development stays out of production.
“Create a staging environment.”
From prompt to working integration
The quick-start skill walks your agent through the same steps as our quick start, in your codebase.
- 1
Plan
The agent inspects your project, confirms the categories, integrations, and region you need, and waits for your approval before changing code.
- 2
Activate in the Sandbox
Integrations are activated in the Sandbox, which returns synthetic data and accepts mock credentials — no vendor developer accounts needed yet.
- 3
Add authorization and the first API call
The agent installs the SDK, embeds the Authorization component, adds a callback route that verifies the signature and stores the connection ID, and makes the first call.
- 4
Verify and sync
It tests the flow end to end, checks the API-call logs, and optionally registers webhooks with signature verification for real-time sync.
What you do: create your workspace and copy your API keys, activate integrations, review the agent's plan and changes, and when you launch, register your app with each vendor and switch to Production.
Skills for every category
After the quick start, your agent loads the skill for your use case: object shapes, sync patterns, and category-specific rules.
Prefer the terminal?
The Unified.to CLI gives agents and developers CRUD on any object, passthrough calls to the vendor API, and unified objects to discover every object, method, and list filter.
brew tap unified-to/cli && brew install unified export UNIFIED_API_KEY=... unified objects crm unified list <connection_id> crm_contact --limit 10
Two MCP servers, two jobs
Core MCP — for your coding agent
Builds and operates your integrations: workspace integrations, connections, webhooks, environments, logs, and docs search.
Unified MCP — for your product's agents
Gives the AI features you ship real-time, authorized access to your customers' SaaS data as callable tools.
Frequently Asked Questions
Any agent that can read a URL or connect to an MCP server: Claude Code, Cursor, OpenAI Codex, GitHub Copilot in VS Code, Windsurf, Zed, Devin, and others. The quick-start skill is plain Markdown, and the Core MCP server uses the standard Streamable HTTP transport.
The Core MCP server is for you, the developer. It lets your coding agent manage your Unified.to workspace (integrations, connections, webhooks, environments, API-call logs, issues) and search the docs. It never reads or writes your end-customers' data.
Unified MCP is for your product's AI agents. It exposes an authorized customer connection's data as callable tools inside LLM APIs.
Agents use the official SDKs for TypeScript, Python, PHP, Java, Go, C#, and Ruby, or plain REST from any language. The embedded Authorization component ships for React, Vue, Angular, Svelte, and plain JavaScript. See SDKs.
A Unified.to workspace, your Workspace ID and API key (Settings → API Keys in app.unified.to), and at least one integration activated in the Sandbox. Everything else — installing the SDK, adding the Authorization component, the callback route, the first API call, and webhooks — the agent can do in your codebase.
Yes. The Sandbox environment returns synthetic data and accepts any mock credentials, so the agent can build and verify the whole flow end to end without vendor developer accounts. When you launch, register your app with each vendor, activate the integration in Production, and switch env=Sandbox to env=Production. See Sandbox.
The workspace API key grants access to your whole Unified.to account, so treat it like a production secret. Pass it to the MCP server from an environment variable or a secure prompt rather than pasting it into a config file you commit, point your agent at the Sandbox while building, and keep the key server-side in your app.
Yes. After the quick start, agents load a category skill for object shapes and patterns — CRM, ATS, HRIS, accounting, payments, e-commerce, messaging, calendar, ticketing, enterprise search/RAG, and more. They are listed in the docs index.
Shipintegrations.Faster.