Partner SuiteNew

Native MCP for the tools your AI already uses

Connect Claude, ChatGPT, Cursor, or any MCP client to a production-grade tool server. Your AI agent discovers Synup tools automatically, calls them with full OAuth 2.1 security, and respects multi-tenant isolation out of the box.

JSON-RPC 2.0 Streamable HTTP OAuth 2.1 + PKCE 3 tool domains
Start free trial Read the MCP docs
Production MCP server, not a preview 14-day free trial · No credit card
Agencies wiring their AI agents into Synup
The protocol

One endpoint. Every tool.

POST /mcp — that's the entire surface. Your AI agent sends a JSON-RPC 2.0 request, Synup executes the tool, and returns the result. The Streamable-HTTP transport means no WebSocket complexity. GET /mcp returns a WWW-Authenticate header so MCP clients can discover the auth server automatically.

Single endpoint: POST /mcp
JSON-RPC 2.0 over Streamable-HTTP — no WebSockets
Auth discovery via GET /mcp WWW-Authenticate header
Multi-tenant isolation: agencyId resolved from auth context, never from tool params
Tool registry

Three tool domains, ready to call

The registry ships with CRM tools (companies, contacts, deals), post idea tools (create, update, archive, publish), and post tools (create, schedule, publish, analyze). Each tool declares its input schema, whether it mutates data, and which scopes it requires.

3tool domains
2scopes
1endpoint
CRM toolscrm/* 4 of 9
ToolScopeMutates
list_companiesCompanies with owner + stagemcp:readNo
list_contactsContacts, filterable by companymcp:readNo
list_dealsDeals with value + stagemcp:readNo
list_stagesPipeline stage definitionsmcp:readNo
Read-only domain — nothing here can write
Post idea toolsmcp/tools/post-ideas 4 of 6
ToolScopeMutates
create_post_ideaDraft an idea against a clientmcp:writeYes
list_post_ideasIdeas by client and statusmcp:readNo
publish_idea_to_postPromote an idea into a postmcp:writeYes
archive_post_ideaRetire an idea from the queuemcp:writeYes
Mutating tools require mcp:write
Post toolsmcp/tools/posts 4 of 7
ToolScopeMutates
create_postSchedule to Google + Facebookmcp:writeYes
list_postsPosts by client, status, windowmcp:readNo
publish_post_nowPush a scheduled post livemcp:writeYes
get_post_analyticsViews, clicks, engagementmcp:readNo
The domain the hero call above uses
View tool input schemas create_post 
{
  "name": "create_post",
  "description": "Create a social post for a client",
  "mutates": true,
  "scopes": ["mcp:write"],
  "inputSchema": {
    "type": "object",
    "required": ["clientId", "name", "platforms"],
    "properties": {
      "clientId":     { "type": "string" },
      "name":         { "type": "string", "maxLength": 120 },
      "platforms":    { "type": "array", "items": { "enum": ["google", "facebook", "instagram"] } },
      "postType":     { "enum": ["announcement", "offer", "event"] },
      "scheduledFor": { "type": "string", "format": "date-time" }
    }
  }
}
// Illustrative schema, swap for the live registry output of tools/list.
Security

OAuth 2.1 — not a shared secret

MCP auth uses the same OAuth 2.1 server as the REST API. Your MCP client registers dynamically (RFC 7591), gets a client_id, runs the PKCE authorization flow, and receives scoped tokens. Three auth paths are supported for different deployment contexts.

Scopes: mcp:read for read tools, mcp:write for mutating tools
Multi-tenant: agencyId comes from the auth context only — tools cannot impersonate
Keys, IDs and headers shown are illustrative samples, not live credentials.

AI-native from the ground up

Four things the server does so your agent doesn't have to.

Streamable HTTP transport

JSON-RPC 2.0 over standard HTTP POST. No WebSocket complexity, no custom protocols.

POST /mcp HTTP/1.1
Content-Type: application/json
Authorization: Bearer eyJhbGciOi…

Tool discovery

MCP clients automatically discover available tools, their schemas, and required scopes. No manual configuration.

// tools/list → 22 tools
{ "name": "list_posts",
  "inputSchema": { "clientId": "string", "status": "enum" },
  "scopes": ["mcp:read"] }

Multi-tenant isolation

Agency ID is resolved from the auth token, never from tool parameters. Tools cannot reach another agency's data.

// decoded access token
{ "agencyId": "agency_pinnacle",
  "scopes": ["mcp:read", "mcp:write"],
  "exp": "+1h" }

Agent container sync

Save an API key or Anthropic key once and it propagates to the agent container, which reloads on its own.

Agent containeragency_pinnacleSynced
// Marinara Pizza — NYC rollout
"objective": "maximize presence, manage reputation",
"market": "NYC — most competitive pizza market in the US"
// result: 80% increase in phone calls
MP Marinara PizzaRestaurants & Food
Marinara Pizza "

Marinara Pizza added Synup to their marketing plan a few months after opening. Working with Synup's Customer Success team, they mapped out a strategy to maximize presence everywhere and aggressively manage reputation from day one — in one of the toughest pizza markets in the country.

Marinara PizzaNYC, multiple locations
Read the full story →
FAQ

Start here if MCP is new to you

What is MCP?
Model Context Protocol (MCP) is an open standard for connecting AI agents to external tools. Instead of writing custom API integrations, your AI agent discovers and calls Synup tools natively through a single endpoint.
Which AI clients work with this?
Any MCP-compatible client: Claude (via Anthropic's desktop app or API), ChatGPT (with plugin/MCP support), Cursor, Windsurf, and custom MCP clients. The protocol is standardized — if it speaks MCP, it works.
What tools are available?
Three tool domains: CRM (companies, contacts, deals, stages), Post Ideas (create, update, archive, publish), and Posts (create, schedule, publish, analyze). More domains are being added.
How is this different from the REST API?
The REST API is for developers building custom integrations with HTTP calls. MCP is for AI agents that discover and call tools autonomously. The REST API requires you to write code; MCP lets your AI agent figure out which tools to call.
Is it secure?
Yes. MCP auth uses OAuth 2.1 with PKCE, the same standard used by the REST API. Tokens are scoped (mcp:read, mcp:write), short-lived (1h), and agency-isolated. Tools cannot access other agencies' data.
How do I get started?
Create an API key in Settings → API Keys, set the Synup MCP URL, and point your MCP client at the endpoint. The client will discover available tools automatically.

Let your AI agent do the work

Point your MCP client at the endpoint, approve the scopes, and start calling tools in the same conversation you're already having.

// add to your MCP client config
{ "synup": { "url": "https://mcp.synup.com/mcp", "auth": "oauth" } }
Start free trial Read the MCP docs
No credit card required Live sync in minutes Cancel anytime