integrations tools snowplow
Snowplow MCP for your AI agent
Cogny's Snowplow integration exposes 20 tools (13 read, 4 write, 3 destructive) to an AI agent over MCP. The same tools are available on every plan, from your own AI client or inside the Cogny app.
- 13readonly reads data.
- 4writecreates or changes something.
- 3destructivedeletes, removes or replaces something.
Over MCP your own AI client asks before it runs a write. To take a tool away from your agent, block it in Settings → Integrations. Plans, approvals and limits
set it up.
Add Cogny to your AI client once, then connect Snowplow in Cogny. Every Snowplow tool below appears in the client.
- Claude CodeRun
claude mcp add --transport http cogny https://app.cogny.com/mcp, then authorise in the browser. - Claude.aiSettings → Connectors → Add custom connector, URL
https://app.cogny.com/mcp, then log in and pick your workspace. - ChatGPTSettings → Apps → Advanced settings → Developer mode → Create app, URL
https://app.cogny.com/mcp, authentication OAuth. - CursorIn Cogny, Settings → Data & Integrations → Connect AI Clients shows the block to paste into
.mcp.json, with an API key. - CodexRun
codex mcp add cogny --url https://app.cogny.com/mcp.
Notes
Snowplow connects with a scoped API key created in the Snowplow Console (Settings → Manage organization → View and manage API keys → Create API key, type "Snowplow API key"). Cogny needs three values: the Organization ID (from the Manage organization page), the API Key ID and the API Key itself. Use a dedicated scoped key rather than a legacy Global admin key. A key works for one organization, so the tools use the connected organization by default.
The connection covers your Snowplow tracking setup: data structures (schemas) and their drafts, tracking plans (called data products in the API), event specifications, pipelines, and the data catalog. The key's View permission decides what can be read; the write tools also need Manage on the feature they change.
Writes that change something already in use ask for confirmation first. update_data_product, update_event_spec, deploy_data_structure, and create_event_spec with a status other than draft return the exact request they would send and change nothing until they are re-run with confirm=True. The update tools keep every field you don't change: they read the current tracking plan or event spec and send it back with only your changes. Creating a schema draft, validating a schema, creating a tracking plan and creating a draft event spec run directly, because they add something new and change nothing existing.
Promoting a schema to production follows Snowplow's own order (validated → development → production), and only keys with admin rights may deploy to production.
Cogny exchanges the key for a short-lived access token on each call and refreshes it automatically. A key that stops working marks the connection as needing a reconnect.
Known gaps
- Not yet run against a live Snowplow organization. Every endpoint, method and request body was checked against Snowplow's published Console API specification and documentation, but the responses have only been tested against samples.
- Two details aren't settled by Snowplow's documentation: whether an event-spec edit that doesn't send
statuskeeps the current status (the update tools send the current status back to be safe), and whether a tracking-plan update needsstatus(the tracking plans guide says yes, the API specification doesn't list it; the current value is sent back). - No raw event data. The Console API only covers the tracking setup; the events themselves live in your own warehouse and are reached through a database connection, not this integration.
- No delete tools. Deleting a tracking plan, event specification or schema draft is done in the Snowplow Console.
- No failed-event metrics, collector configuration, enrichment settings, source applications or tracking-plan subscriptions yet, although the Console API has endpoints for them.
- No heatmaps, recordings or session replays: the Console API doesn't offer them.
what each tool does.
Read 13 · only reads data
list_data_structuresReadList the data structures (schemas) in your Snowplow organization, with where each is deployed.get_data_structureReadGet one data structure (schema) and its deployments.get_data_structure_versionReadGet the JSON schema of one version of a data structure.list_data_structure_draftsReadList the data structure drafts (schemas not yet deployed) in your Snowplow organization.get_data_structure_draftReadGet one data structure draft.list_data_productsReadList the tracking plans (data products) in your Snowplow organization.get_data_productReadGet one tracking plan (data product), with its event specifications.list_event_specsReadList the event specifications in your Snowplow tracking plans.get_event_specReadGet one event specification, with any schema-compatibility warnings.list_pipelinesReadList your Snowplow pipelines (production and development environments).get_pipelineReadGet one Snowplow pipeline: its name, status, label and configuration.list_data_catalogReadList the events and entities in your Snowplow data catalog, and which are tracked together.search_data_catalogReadSearch your Snowplow data catalog for an event or entity by name or description.
Write 4 · creates or changes something
create_data_structure_draftWriteCreate a draft of a new data structure (schema) in Snowplow.validate_data_structureWriteCheck that a data structure (schema) is valid JSON Schema and loads into your warehouse.create_data_productWriteCreate a new tracking plan (data product) in Snowplow.create_event_specWriteAdd a new event specification to a Snowplow tracking plan.
Destructive 3 · deletes, removes or replaces something
deploy_data_structureDestructiveDeploy a schema version to a Snowplow environment, e.g. from development to production, only after the user confirms.update_data_productDestructiveChange the name, description, domain, owner or source apps of an existing tracking plan, only after the user confirms.update_event_specDestructiveEdit, publish or deprecate an existing event specification, only after the user confirms.
Exact input schemas come from tools/list once Snowplow is connected; the free plan is enough to check. Missing something? Your agent can ask for it with request_feature.