Dreamwork

What the Dreamwork MCP server does

The Dreamwork MCP server gives AI assistants access to Dreamwork’s job-search and application workflow. It can find ranked job matches, inspect full listings, save or dismiss roles, and maintain a pipeline of saved and applied positions. It also supports importing public job URLs into a private pipeline and recording applications submitted outside Dreamwork.

For application preparation, the server can queue tailored resumes, cover letters, and employer-question answers. Agents can inspect preparation status, edit or strengthen existing materials, change resume formatting, restore earlier versions, and retrieve short-lived document download links. Application records can be moved through board columns, and queued applications can be cancelled when submission is still preventable.

How it works

The Dreamwork MCP server is a gateway rather than a database-backed local application. The client sends requests to the hosted Dreamwork API over HTTPS. The published package runs as a stdio MCP process, with JSON-RPC data written to stdout; a Streamable HTTP endpoint is also available at https://mcp.dreamworkhq.com/mcp.

Without credentials, public browsing is available in guest mode. Authenticated account tools use a Dreamwork API key or hosted sign-in and can access account-specific matches, preferences, materials, pipeline data, and application actions. Consequential operations use a confirmation step in the assistant workflow before proceeding.

Setup and configuration

Run the stdio client on demand with:

npx -y @dreamworkhq/mcp

Claude Desktop and Cursor can launch that command from their MCP configuration. Claude Code can use the hosted endpoint with an authorization header. The hosted endpoint can also be added as a custom connector in supported products such as Claude or ChatGPT.

DREAMWORK_API_KEY is optional for starting the server but is needed for account features. DREAMWORK_API_URL can override the default API endpoint, and DREAMWORK_UNREAD_NOTICES controls unread recruiter-mail notices attached to tool results. Telemetry can be disabled with DREAMWORK_TELEMETRY or DO_NOT_TRACK.

Tools and capabilities

The Dreamwork MCP server covers these workflows:

• Search and paginate ranked matches, inspect job details, and filter discovery windows.
• Save, dismiss, restore, and import roles without confusing pipeline changes with employer actions.
• Read and update preferences, parse preference text, and review application readiness.
• Prepare application packs, check generation status, edit materials, replace documents, and download versioned files.
• Record off-site applications with person attestation and track pipeline columns, interviews, and application state.
• Store career facts and hypotheses separately, then confirm hypotheses when the person verifies them.

The tools distinguish between recording a user-reported state and contacting an employer. Moving a card to Applied does not submit an application, and marking an off-site application does not send one.

Limitations and notes

A Dreamwork account and API key are required for matches ranked against a resume, saved jobs, tailored materials, pipeline management, and applying. Public guest access is read-only. API-key permissions and plan allowances can limit available operations.

Applications already submitted cannot be recalled. Materials for locked or sent applications cannot be replaced, and application edits require existing materials. Eligibility may be unknown; only a confirmed restriction is treated as blocking by the job details workflow. Download links for application documents are temporary and typically remain usable for about ten minutes.

The server does not provide its own Dreamwork database, API, workers, or hosted OAuth service. A stdio wrapper must not print banners to stdout because that can corrupt the MCP JSON-RPC stream.