MCP Server

NearSync runs a hosted Model Context Protocol (MCP) server so an AI agent — Claude, ChatGPT, an MCP-capable IDE — can read your data in natural language ("how's my pipeline?", "what does my day look like?", "do I have the call with Acme today?").

It is read-only, org-isolated, and permission-aware: an agent only ever sees data the connecting user is allowed to see. There are two surfaces:

Org analytics — pipeline, revenue, MRR, etc. — gated by your role (a sales rep can't read finance metrics).
Personal — your tasks and schedule — scoped to you.

Endpoint

https://api.nearsync.tech/mcp

Streamable HTTP (JSON-RPC 2.0). Two ways to authenticate:

Option A — Connect with OAuth (recommended)

One click, no keys to copy. Most MCP clients let you add a remote server by URL and run the OAuth flow themselves (Dynamic Client Registration + PKCE — nothing to pre-register):

In your client, add a custom/remote MCP connector with the URL https://api.nearsync.tech/mcp.
The client opens a NearSync consent screen — sign in (or you're already signed in), confirm which account + org, and click Allow.
Done. The agent now has the tools, scoped to your permissions.

The agent only sees:

the analytics your role permits (e.g. finance metrics only if you have Finance access), and
your own tasks and schedule.

Tokens are short-lived and refresh automatically; revoke a connection any time from your client.

Option B — Connect with an API key (developer / automation)

For scripts, automation, or clients without OAuth. Mint a key in System → API Keys (needs the analytics:read scope, and tasks:read/schedule:read for personal tools), then pass it as a bearer token. For clients that bridge stdio to a remote server (e.g. Claude Desktop):

{
  "mcpServers": {
    "nearsync": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://api.nearsync.tech/mcp",
               "--header", "Authorization: Bearer ns_live_..."]
    }
  }
}

Keys vs OAuth

A key is a machine credential — it carries whatever scopes the admin granted and is not narrowed per-user (use it for automation an admin owns). OAuth is per-user and RBAC-narrowed — the agent sees only what that person is cleared for. Personal tools (my_tasks, etc.) require OAuth (a key has no user).

Tools

Tool	Args	Returns	Scope
`list_metrics`	`category?`	The metrics/reports you can read (filtered to your permissions). Start here.	`analytics:read`
`run_metric`	`metric` (name or id), `period?`, `filters?`	One metric's result (number / series / table / computed). Denied if your role can't see its category.	`analytics:read`
`dashboard_insights`	—	"What changed" — metrics with material movement (stale-data drops flagged, not counted as declines).	`analytics:read`
`org_mrr`	—	Monthly recurring revenue (finance-gated).	`analytics:read`
`org_health`	—	Organization health score.	`analytics:read`
`my_tasks`	`filter?` (open/overdue/today/upcoming/all)	Your tasks.	`tasks:read`
`my_schedule`	`date?` / `range?`	Your meetings/calls.	`schedule:read`
`my_day`	`date?`	Your tasks + meetings for a day.	`tasks:read`

Tools appear in tools/list only when your connection carries the scope. run_metric accepts a metric name (from list_metrics) or its UUID — the server resolves names. Money is in your organization's base currency.

Example

{ "jsonrpc": "2.0", "id": 1, "method": "tools/call",
  "params": { "name": "my_day", "arguments": { "date": "2026-06-03" } } }

Limits & errors

401 — missing/invalid/expired credential (re-auth). 403 / tool isError — your role lacks the metric's category, or the tool needs a connected user (personal tools over a key). API keys count against a per-minute rate limit + monthly quota (see Rate Limits & Quota); quota is charged only on a call that reaches data.

Endpoint​

Option A — Connect with OAuth (recommended)​

Option B — Connect with an API key (developer / automation)​

Tools​

Example​

Limits & errors​