Origin
by Origin
Query your organisation's own AI-agent telemetry in natural language. Two tools — describe the analytics graph, then run a structured query against it — with role-based access, a per-query cost budget, and every MCP tool call written to an audit log Origin documents in detail.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Connect Origin via MCP
https://mcp.prod.originhq.com/mcpWorks in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.
Origin Tools & Capabilities (2)
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
Limits
- The read grant is tenant-wide, whichever role you use. Origin's capability matrix gives Admin and Member identical visibility into endpoints, events, prompt sessions and telemetry. Role choice does not narrow what an agent can query.
- The data is employee AI behaviour. Prompts, sessions and per-user activity attributed to people and machines. A read-only label does not make bulk export uninteresting.
- The schema tool hands over the full map.
analytics_describe_schemareturns every entity, field, metric and join edge, so the model knows the whole queryable surface before its first query. - Audit retention is 30 days. Detailed, per-tenant, and short. Export if your review cycle is longer than a month.
- No documented row cap.
limitand atruncatedflag bound each payload, but Origin publishes no default or maximum row count for a query. - Plan query limits are unquantified. Origin's pricing names capped queries on the free tier and a monthly cap or credit pack on Team without stating any figure.
- The cost budget is per query, not per session. It rejects one oversized query; it does not stop a sequence of small ones adding up to the same export.
- No tool annotations are published. Neither tool carries a machine-readable read-only hint, and the gated endpoint meant we could not check. The read-only claim is prose.
- OAuth scopes describe identity, not access.
openid,profile,emailandoffline_accesssay nothing about the analytics grant a user is approving. - Everything about enforcement is documented, not observed. We did not authenticate, so the role-scoping, the audit log and the cost budget are Origin's stated behaviour as of 2026-08-19.
- Your data lands in the client. Query results arrive in whatever MCP host you connected, with whatever retention that host has.
- The support site blocks crawlers.
support.originhq.com/robots.txtdisallows all user agents, so the documentation cited here is reachable but not indexed. - Identity integrations are partly unshipped. Origin lists Entra ID as available, with Google ID and Okta both marked as coming soon.
Frequently asked questions
What data does the Origin MCP server actually query?
Your own organisation's AI activity, not a customer or finance dataset. Origin is an endpoint AI observability platform that inventories what AI agents do on company laptops. The MCP server exposes the same analytics graph behind Origin's dashboard: endpoints, AI prompts, sessions, models called, token spend, and tool-call chains, attributed to users and processes.
Does analytics_query accept raw SQL?
No. Origin documents a structured query object instead: an entity, joins, filters, a select list, metrics, ordering, paging, a time range and an optional sample. The model composes fields into that shape rather than writing SQL text, and malformed queries come back as validation errors. That containment is real and materially narrows the surface of a broad read grant.
Does Origin log MCP queries so an administrator can review them?
Yes, and in unusual detail. Origin states every tool call is recorded to its audit log with the calling user, the tool name, a hash and snapshot of the arguments, the row count, bytes scanned, elapsed time and outcome. Records are scoped per tenant and retained for 30 days. This is the strongest audit posture in this catalogue.
Does Origin's role-based access control apply over MCP?
Yes. Origin documents MCP queries as running with your user identity, and states you can only see what your Origin role allows in the dashboard. The roles are Admin and Member; both can read all tenant telemetry, so the practical grant for either role is the whole tenant's AI activity. Role separation governs configuration, not read scope.
How many tools does the Origin MCP server have?
Two, and Origin's documentation agrees with Anthropic's directory snapshot exactly. The vendor names `analytics_describe_schema` and `analytics_query`, states the server exposes two tools, and describes it as read-only with no ability to modify endpoints, agents or alerts. That matching count is rarer in this catalogue than it should be.
What limits a runaway Origin analytics query?
A tenant cost budget. Origin states queries exceeding it are rejected with a suggestion to narrow the scope, and that every response carries a cost object reporting rows and bytes scanned plus elapsed time. Results also carry a truncated flag, the time range defaults to 30 days, and requests are capped at a 256 KiB body.
Why does the Origin MCP endpoint have prod in the hostname?
Because Origin names its production estate that way throughout. The agent installation guide lists two more production hosts under the same convention. For a reader the practical value is unambiguity: the listed endpoint states plainly which environment you are connecting to, rather than leaving you to infer it from a bare vendor domain.
Do you need a paid Origin plan to use the MCP server?
Not necessarily. Origin publishes a free Personal tier for one seat with 14-day retention and, in its own words, capped queries. Paid Team and Team Pro tiers extend retention to 30 and 90 days and add a monthly query cap or credit pack. The MCP server is not listed as a plan-gated feature, so entitlement follows your data access.
Sources
- Origin MCP Server — https://support.originhq.com/docs/origin-mcp-server (fetched 2026-08-19 as Markdown via the
.mdsuffix) · retrieved 2026-08-19 - What is Origin? — https://support.originhq.com/docs/what-is-origin (fetched 2026-08-19) · retrieved 2026-08-19
- Welcome to Origin — https://support.originhq.com/docs/welcome-to-origin (fetched 2026-08-19) · retrieved 2026-08-19
- Role Based Access Control — https://support.originhq.com/docs/role-based-access-control (fetched 2026-08-19) · retrieved 2026-08-19
- Chat Interface — https://support.originhq.com/docs/chat-interface (fetched 2026-08-19) · retrieved 2026-08-19
- Identity Integration — https://support.originhq.com/docs/identity-integration (fetched 2026-08-19) · retrieved 2026-08-19
- Prerequisites & Compatability — https://support.originhq.com/docs/prerequisites-and-known-issues (fetched 2026-08-19) · retrieved 2026-08-19
- Origin pricing — https://www.originhq.com/pricing (fetched 2026-08-19; four tiers, free Personal seat, no per-seat figures published for paid tiers) · retrieved 2026-08-19
- Prelude is now Origin — https://www.originhq.com/blog/prelude-is-now-origin (fetched 2026-08-19) · retrieved 2026-08-19
- Origin docs machine index — https://support.originhq.com/llms.txt (fetched 2026-08-19); marketing-site index at https://www.originhq.com/llms.txt (fetched 2026-08-19) · retrieved 2026-08-19
support.originhq.com/robots.txtdisallows all user agents;www.originhq.com/robots.txtallows all crawlers except a/lp/path and names AI crawlers individually. Neither publishesContent-Signalaxes.support.originhq.com/sitemap.xmlreturns HTTP 404 (all fetched 2026-08-19) · retrieved 2026-08-19- Live anonymous probe — 2026-08-19: an
initializePOST tohttps://mcp.prod.originhq.com/mcpreturns HTTP 401 withwww-authenticate: Bearer realm="origin-mcp", resource_metadata="https://mcp.prod.originhq.com/.well-known/oauth-protected-resource/mcp"· retrieved 2026-08-19 - Live RFC 9728 check — all three protected-resource path forms probed 2026-08-19: root and path-append return identical 155-byte JSON descriptors naming
https://auth-web.dashboard.originhq.com/as the authorization server; path-insert returns a 77-byte JSON unauthenticated error with HTTP 401 · retrieved 2026-08-19 - Live RFC 8414 check — https://mcp.prod.originhq.com/.well-known/oauth-authorization-server returns HTTP 404 with a
not_foundbody at both root and path-append forms (fetched 2026-08-19) · retrieved 2026-08-19 - Live authorization-server metadata — https://auth-web.dashboard.originhq.com/.well-known/oauth-authorization-server (fetched 2026-08-19;
S256only,scopes_supported: ["email", "offline_access", "openid", "profile"], Dynamic Client Registration endpoint present) · retrieved 2026-08-19 - Anthropic connector directory — https://claude.ai/directory/4add6101-9562-4ebb-9ac7-ebdaabed6257 (snapshot 2026-08-16) · retrieved 2026-08-16
Use in Agentman
Connect once and your agents call these tools on their own — on a schedule, in a workflow, with nobody at the keyboard.
Open in Agentman StudioServer Info
- Category
- Data & Research
- Developer
- Origin
- Tools
- 2
- Domain
- mcp.prod.originhq.com
Using Claude Desktop or another MCP client? Setup docs — the connection URL above works anywhere.