Origin MCP Server: Query Observability Data from AI Tools

Origin’s Model Context Protocol (MCP) server lets you query your endpoint and agent observability data from any MCP client that supports remote servers — Claude (desktop and web), Claude Code, Cursor, Cline, and others. Ask natural-language questions about your AI agent activity, prompt usage, model spend, and endpoint behavior, and the LLM you’re using translates them into structured queries against your Origin tenant. This page covers what the server does, how to connect, sample prompts, common errors, and how to get support.

What the MCP server gives you

Origin’s MCP exposes read-only access to your analytics graph — the same data that powers the Origin console. From an MCP client you can answer questions like:

Which agents are running on which endpoints, and what models are they calling?
How much are we spending on each model, broken down by department or team?
What prompts fail most often, and where?
Which sessions had the longest tool-call chains yesterday?

The server is read-only. It cannot modify endpoints, agents, alerts, or any other Origin state.

Connect a client

The server is a remote MCP server. It speaks Streamable HTTP and uses OAuth 2.1 with Dynamic Client Registration, so any MCP client that supports remote HTTP servers with OAuth can connect.


Server URL	`https://mcp.prod.originhq.com/mcp`
Transport	Streamable HTTP
Authentication	OAuth 2.1 via your Origin login (WorkOS AuthKit)

The first time you connect, your client opens a browser window so you can sign in to Origin. After that, your client refreshes tokens automatically.

You do not need to provision API keys, copy bearer tokens, or paste anything from the Origin console. The OAuth flow handles everything.

Claude (desktop and web)
Claude Code
Cursor
Cline
Other clients

In the Claude apps, remote MCP servers are added as custom connectors. Custom connectors are available on Pro, Max, Team, and Enterprise plans.

Do not add Origin to claude_desktop_config.json. That file only supports local servers that Claude launches with a command. A url entry there will not connect, and depending on your Claude version it can stop the app from loading your other MCP servers.

Pro and Max plans

Open Settings → Connectors.
Click Add custom connector.
Enter a name (for example, Origin) and the server URL: https://mcp.prod.originhq.com/mcp. Leave the Advanced settings OAuth fields blank.
Click Add, then click Connect and sign in with your Origin account in the browser window that opens.

Team and Enterprise plans

A Claude organization Owner adds the connector once in Organization settings → Connectors → Add custom connector, using the URL above.
Each member then opens Settings → Connectors, finds Origin in the list, clicks Connect, and signs in with their own Origin account.

Once connected, enable Origin from the + (tools) menu in a chat if it isn’t already on, then ask a question.

Custom connectors connect from Anthropic’s cloud, not from your machine. This works with Origin’s production server, which is publicly reachable.

Add the server with the CLI:

claude mcp add --transport http origin https://mcp.prod.originhq.com/mcp

By default this adds the server for the current project only. Add --scope user to make it available in every project.Then start claude, run /mcp, select origin, and choose Authenticate to sign in with your Origin account in the browser.

Add Origin to ~/.cursor/mcp.json (all projects) or .cursor/mcp.json in a project root (that project only):

{
  "mcpServers": {
    "origin": {
      "url": "https://mcp.prod.originhq.com/mcp"
    }
  }
}

Do not add an auth block or headers. Cursor registers itself with Origin automatically and prompts you to sign in.Open the MCP section of Cursor’s settings, confirm origin appears and is enabled, and click Connect or Login if Cursor shows that it needs authentication.

In the Cline panel, click the MCP Servers icon and open the Remote Servers tab.
Enter a server name (for example, origin) and the server URL: https://mcp.prod.originhq.com/mcp.
Set Transport Type to Streamable HTTP.
Click Add Server, then complete the sign-in prompt.

If you edit Cline’s MCP settings file directly, set type explicitly. Without it, Cline uses the legacy SSE transport, which Origin does not support.

{
  "mcpServers": {
    "origin": {
      "type": "streamableHttp",
      "url": "https://mcp.prod.originhq.com/mcp"
    }
  }
}

Any client that supports remote MCP servers over Streamable HTTP with OAuth can connect. Point it at https://mcp.prod.originhq.com/mcp, choose the Streamable HTTP (sometimes labeled “HTTP”) transport rather than SSE, and follow your client’s instructions for OAuth sign-in.If your client only supports local (stdio) servers, it cannot connect to Origin directly.

Tools exposed

Your client discovers these tools automatically once connected. You don’t call them yourself; the LLM picks the right one based on your question.

Schema and query tools

The LLM typically uses these in order: read the map, pull details for the entities it needs, check the query grammar, then run the query.

Tool	What it does
`analytics_describe_schema`	Returns a compact map of the analytics graph: entities, how they join, and which data domains exist. With no arguments it returns the core domain (endpoints, tokens, inventory, identity). Pass `domains` with one of `conversation`, `os_telemetry`, or `legacy` to see that domain’s map. Field lists and examples are not included.
`analytics_describe_entities`	Takes a list of entity names from the schema map (for example `["ENDPOINT", "TOKEN"]`) and returns their fields, metrics, and worked query examples.
`analytics_query_guide`	Returns the query grammar and common pitfalls. Takes no arguments.
`analytics_query`	Runs a structured query against your tenant’s data and returns columns, rows, a `truncated` flag, and cost details.

analytics_query parameters (the LLM fills these in for you):

Parameter	Purpose
`entity`	Required. The root entity, using names from the schema map (for example `ENDPOINT`, `TOKEN`, `AGENT_INVENTORY`)
`user_intent`	Required. A short summary of what you’re trying to find out, recorded with the query
`alias`	Reference name for the root entity within the query
`joins`	How to traverse into related entities
`filters`	Predicates applied before aggregation
`select`	Dimension fields to return (optionally bucketed by time)
`metrics`	Aggregations to compute (counts, sums, averages, percentiles)
`order_by`, `limit`, `offset`	Sort and paging. `limit` defaults to 500
`time_range_days`	Defaults to 30. Capped at each entity’s data retention
`sample`	Distribution-preserving sampling for broad questions. Can’t be combined with `metrics` or `offset`

Queries that exceed your tenant’s cost budget are rejected with a suggestion to narrow the scope.

Conversation tools

Tool	What it does
`analytics_read_conversation_record`	Reads a page of messages from one agent conversation, using a trace ID returned by `analytics_query`.
`analytics_read_conversation_content`	Reads specific pieces of a conversation in full, such as one message part, tool call arguments, or tool results.

Prior work search

Tool	What it does
`search_prior_work`	Finds how similar work was previously requested and handled in your tenant’s agent activity, so the LLM can build on it.

Sample prompts

Once connected, you can ask questions in plain English. Examples to try:

“What models are my Claude Code agents calling this week, and on which endpoints?”
“Show me total token spend by department over the last 30 days, sorted highest first.”
“Which prompts are failing most often? Group by model and tool name.”
“Find the ten longest agent sessions yesterday and show me which tools each one called.”
“Compare prompt latency p50 and p95 across our models for the past two weeks.”
“Which endpoints had AI activity for the first time this week?”
“How many Claude Code sessions did each engineer run this week?”
“Show me MCP server connections by endpoint — which agents are connecting to which servers?”

For the best results, give the LLM enough specificity to compose a query: include a time range, a grouping dimension, and what you want to measure. If the LLM asks clarifying questions, answer them — it’s usually trying to disambiguate between similar fields.

Troubleshooting

I added the server but no Origin tools appear in my client

First, check how you added it:

Claude desktop or web: Origin must be added under Settings → Connectors, not in claude_desktop_config.json. If you added a url entry to that file, remove it, fully quit and reopen Claude, and add Origin as a custom connector instead.
Cline: Make sure the transport is Streamable HTTP ("type": "streamableHttp"), not SSE.
All clients: Confirm the URL is exactly https://mcp.prod.originhq.com/mcp and that you completed sign-in.

If it still doesn’t work, fully quit and reopen the client and check its MCP logs for connection errors.

I can't find Add custom connector in Claude

On Team and Enterprise plans, only organization Owners can add custom connectors. Ask an Owner to add Origin in Organization settings → Connectors. Members can then connect from Settings → Connectors. Custom connectors aren’t available on the Free plan.

Your token may have expired and your client may not be refreshing it. Disconnect and reconnect the server in your client’s settings. If the problem persists, your Origin account may not have access to the tenant you’re trying to query — confirm in the Origin console that you can see the data there first.

Validation error / DSL query errors in responses

The LLM composed a query the server couldn’t validate — usually a field referenced on the wrong entity, or a filter value of the wrong type. These come back as visible errors so the LLM can self-correct on the next turn. If it loops, ask the LLM to call analytics_describe_entities for the entities it’s querying and analytics_query_guide before retrying.

Cost exceeds budget errors

You asked for something that would scan more rows than your tenant’s per-query budget allows. The error message includes a suggestion — usually: narrow the time range, add a filter, or sample the result. Time-bound your question more tightly and try again.

Empty results when you expected data

The LLM may have applied a filter that doesn’t match anything. Ask it to show the query it ran and check the filters. Common culprits: a hostname filter with the wrong case, or a time range that excludes the data you care about. Some entities keep data for a limited time, so very old data may no longer be available.

Slow responses

Analytics queries scale with the data they scan. If a question consistently takes more than 10–20 seconds, try narrowing the time range or asking for an aggregate (counts, top-N) instead of raw rows.

Versioning and protocol


MCP protocol version	`2025-06-18` (with fallback to `2025-03-26` and `2024-11-05`)
Transport	Streamable HTTP (POST `/mcp`)
Auth	OAuth 2.1 with Dynamic Client Registration (RFC 7591) and Protected Resource Metadata (RFC 9728)
Request body limit	256 KiB

Discovery metadata is published at https://mcp.prod.originhq.com/.well-known/oauth-protected-resource for clients that need it. You should not need to configure this manually.

Data, privacy, and audit

Every tool call is recorded to Origin’s 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. Audit records are scoped per tenant and retained for 30 days.

Queries run with your user identity — you can only see what your Origin role allows you to see in the console. See Access Control.

Get support

In-app: Open the support panel in the Origin console.
Email: support@originhq.com

When reporting an issue, include your MCP client and version, the prompt you ran, and any error message the client returned. If the server returned a structured error, the message is usually enough to pinpoint the problem.

Overview

Setup and Configuration

AI Provider Integration

Agent Installation

MDM Deployment

Origin MCP Server: Query Observability Data from AI Tools

What the MCP server gives you

Connect a client

Tools exposed

Schema and query tools

Conversation tools

Prior work search

Sample prompts

Troubleshooting

Versioning and protocol

Data, privacy, and audit

Get support

​What the MCP server gives you

​Connect a client

​Tools exposed

​Schema and query tools

​Conversation tools

​Prior work search

​Sample prompts

​Troubleshooting

​Versioning and protocol

​Data, privacy, and audit

​Get support

What the MCP server gives you

Connect a client

Tools exposed

Schema and query tools

Conversation tools

Prior work search

Sample prompts

Troubleshooting

Versioning and protocol

Data, privacy, and audit

Get support