Skip to main content

Documentation Index

Fetch the complete documentation index at: https://datost.com/docs/llms.txt

Use this file to discover all available pages before exploring further.

Connect any MCP-compatible server to Datost and its tools become callable by the AI analyst inside Slack, alongside your warehouse and business-tool integrations.

​
What is MCP?

The Model Context Protocol is an open standard for exposing tools, resources, and prompts to AI models over a uniform interface. Any HTTP MCP server — vendor-hosted or self-built — can be plugged into Datost without custom integration work.

​
What it unlocks

Adding an external MCP server mounts its tools into Datost’s agent loop. When a question benefits from those tools, the model picks them the same way it picks SQL, PostHog, or HubSpot actions.
  • Reach systems like ticketing, CRM, and observability tools that Datost does not natively support.
  • Wrap proprietary APIs as tools your team can invoke from Slack by mentioning @Datost.
  • Use any vendor that ships a hosted MCP endpoint — Linear, Sentry, Notion, and others.

​
Connect an MCP server

1

Open MCP Servers

In the Datost web app, go to Settings → MCP Servers and click Add server.
2

Configure the endpoint

Give it a name, pick the transport (http streamable is recommended; sse is supported), and paste the HTTPS URL. HTTP is blocked in production; private and loopback hosts are rejected by the server-side URL validator.
3

Add authentication

Choose either Headers (static key/value pairs, typically an Authorization: Bearer ... header) or OAuth 2.1. OAuth auto-discovers the authorization, token, and dynamic-client-registration endpoints, then walks you through a standard consent flow. Secrets are encrypted at rest.
4

Test and list tools

Hit Test connection to run an initialize handshake, then List tools to fetch the server’s advertised tools. Toggle individual tools off to hide them from the agent.

​
Scope to a Slack workspace

MCP servers are defined at the org level but mounted per Slack connection, so a finance workspace and an engineering workspace can see different tool sets.
1

Open a Slack connection

From Settings → Slack, pick the workspace you want to configure.
2

Assign servers

Under MCP Servers, check the servers that should be mounted for that workspace. Up to 25 servers can be assigned per connection.
A server must be enabled at the org level and assigned to the connection for its tools to appear in that Slack workspace.

​
Security considerations

MCP servers run arbitrary tools that the agent can invoke on your behalf. Treat them with the same care as any outbound integration: only connect servers you operate or fully trust, and review the advertised tool list before enabling.
  • Every create/update/test/delete action is written to your audit log.
  • URLs are validated before save and test. Private IP ranges, localhost, and private-network hostnames are blocked.
  • Headers and OAuth tokens are encrypted with per-record keys.
  • Disable individual tools from a server if you want to keep the connection but narrow the blast radius.

​
Limitations

  • HTTPS-only in production.
  • Tool discovery (tools/list) requires HTTP transport; SSE servers can be tested but their tools must be managed from the server side.
  • Maximum 25 MCP servers assigned per Slack connection and up to 200 disabled tools per server.
  • Datost mounts tools only — MCP resources and prompts are not surfaced to the agent today.
Building your own MCP server? Any framework that speaks the 2025-06-18 MCP spec over streamable HTTP will work. Start with a single tool, connect it, and iterate from the Datost tool list view.