42actions
sources, destinations, connections, reverse ETL, syncs & status
Daspire MCP turns your workspace into a Model Context Protocol server. Point Claude Code, Cursor, Codex, or ChatGPT at it and add sources, wire destinations, configure reverse ETL, and trigger syncs — by describing what you want, not clicking through forms. Every write is previewed and confirmed before it runs.
MCP isn't a chatbot bolted onto a dashboard. It's a typed set of workspace tools your coding agent can call directly — so the thing that reads your code can also stand up the pipeline that feeds it. Here's what that buys you in practice.
Daspire exposes every workspace operation as a typed MCP tool with a clear schema. Your agent reads the intent, chooses add_source over create_destination, fills the arguments, and reports back — no SDK to learn, no REST to wire.
Add a source, wire a destination, map streams, and set cadence — each as a discrete tool call the agent composes into one connection. Credentials are handed off through a secure OAuth prompt; secrets never touch the transcript.
Activation is just another tool. Tell the agent which model and which destination tool, and Daspire syncs the rows back to Klaviyo, HubSpot, or any of 200+ operational targets — with field mapping the agent proposes and you approve.
Tokens are workspace-scoped and short-lived. Reads run freely; any create, update, or delete returns a plan the agent must surface for your approval before it executes — and every confirmed action lands in the audit log with who, what, and when.
Daspire MCP follows the open Model Context Protocol, so any compliant host can drive your workspace — no proprietary plugin, no lock-in. Drop one config block into the client you already use and the Daspire tools show up in the next session.
Setting up a pipeline shouldn't mean leaving the editor, clicking through a dozen console screens, and copying IDs between tabs. Daspire MCP lets the agent you're already working with do the clicking — and surface a plan you approve before anything changes.
Add the Daspire MCP server to your client of choice, authenticate once through a scoped OAuth prompt, and the full tool set shows up in your next session. Same server, same tools — whichever host you point at it.
Daspire MCP is built for teams that get audited. Every token is scoped, every write is gated, and every call is logged — so an autonomous agent can be useful without being dangerous.
Auth through OAuth or a scoped PAT. A token reaches one workspace, expires on a schedule, and is revocable in a click.
Write tools return a plan with the blast radius; nothing executes until you approve it in the client. Read-only mode locks writes off entirely.
Each MCP action lands in the workspace audit trail with the actor, timestamp, and payload. Export it or stream it to your SIEM.
Tools respect the same roles and rate limits as the console. Grant an agent analyst access and it simply can't reach admin actions.
If yours isn't here, ask in chat — we don't gatekeep technical conversations behind a sales call.
The full workspace operation set, exposed as 42 typed tools: add and pause sources, create and test destinations, build connections with stream and schedule mapping, configure reverse ETL / activation syncs, trigger and backfill runs, and read status, logs, and MAR usage. Reads run freely; anything that changes state returns a plan you confirm first.
Any host that implements the Model Context Protocol. We test against Claude Code, Cursor, Codex CLI, ChatGPT, Windsurf, Zed, Cline, Continue, VS Code, and Goose — and because it's the open spec, your own MCP host works too. The server runs over stdio locally or streamable HTTP/SSE for remote setups.
No. The server is read-only by default, and every create, update, or delete returns a plan with its downstream impact that the client must surface for your explicit approval before it runs. Nothing executes silently, and every confirmed action is written to the audit log.
You connect with OAuth or a scoped personal access token. Tokens are bound to a single workspace, short-lived, and revocable, and they carry the same role and rate limits as a console seat — so an agent can never exceed the access of the user that authorized it.
No. MCP tools operate on workspace configuration and metadata — connections, schedules, status. Your actual commerce rows move only between your sources, Daspire, and your warehouse, exactly as they do without MCP. The agent sees tool results, not your tables.
No — MCP drives whatever lives in your workspace. It pairs naturally with Extract & Load and Transform (the agent can stand up sources and kick off models for you), but it works against any connections you already have.
14-day free trial — no credit card. Add Daspire MCP to Claude, Cursor, or Codex, authenticate once, and stand up your first connection by describing it.