davinci-resolve-mcp is a Model Context Protocol (MCP) server for DaVinci Resolve Studio that lets AI assistants control Resolve through the official scripting API. It provides full API coverage along with guarded workflow assistance features, and can be used for editing, media pool organization, render setup, review markers, color grading, Fusion, Fairlight, project lifecycle tasks, extension authoring, and source-safe media analysis.
Main capabilities: The server provides both composite tools and fine-grained tools; the composite server is recommended unless a per-method tool surface is explicitly needed. Each composite tool return includes an _operation block alongside the result, containing status (success/partial/blocked/failed), verification, changes, warnings, and execution_id, where verification distinguishes contradictions (Resolve reports success but the read-back result is inconsistent). Multi-step AI operations can be correlated across tool calls into a unified execution trace, aggregating tool duration, call counts, cumulative semantic deltas, and read-back verification, and provides tools such as get_execution_trace, list_recent_executions, begin_execution/end_execution, and export_execution_report, which by default outputs to logs/execution-reports/<execution_id>.md; inspect_operation can assess risk level, destructive potential, and scope of impact before execution.
Usage: Before connecting, DaVinci Resolve Studio needs to be open, and Preferences > General > External scripting using must be set to Local. The npm launcher installs a managed copy under the user application data directory, then runs a general Python installer; the installer creates a virtual environment, detects the Resolve path, and can configure Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Zed, Continue, Cline, Roo Code, OpenCode, Codex CLI, and JetBrains IDEs. Source installation is also supported. The installer and MCP client use ~/.config/davinci-resolve-mcp/bridge.json by default, and the location can be changed via DAVINCI_RESOLVE_BRIDGE_CONFIG. Setting DAVINCI_RESOLVE_BRIDGE=1 forces the bridge transport to be used exclusively.
Local control panel: From the repository root, a single-user local control panel can be launched; it listens only on the loopback address and opens a URL in the browser with a per-launch access token (in the form http://127.0.0.1:8765/#token=…），必须使用该确切 URL, otherwise the panel rejects the request. Persistent analysis tasks automatically refresh the local search index after a successful slice; manual Build Index is used to rebuild from existing reports.
Free version notes: Blackmagic restricts external scripting to the Studio version. On the free version, scriptapp("Resolve") rejects external processes regardless of preferences. As of Resolve 21.0.x, the Workspace ▸ Scripts menu is not restricted, so the server can reach the free version through a small script running inside Resolve, via an authenticated loopback listener. Resolve 21.1 moved Python scripting to Studio, and the Scripts menu in the free version of 21.1 no longer lists .py files, so the bridge should currently be regarded as a 21.0.x path. The bridge is loopback-only, requests are HMAC-signed, and nonces are single-use.
Advanced server (optional, Node): The same package also provides the optional davinci-resolve-advanced-mcp server, which can read and edit Resolve files (.drp/.drt/.drx) and apply database/XML-level changes without Resolve running, and can run in the cloud or locally. 18 tools in total: drp, drt, drx, offline_ref, conform, color_trace, fusion, audio_plan, fairlight, audio, project_read, project_db, pipeline, capabilities, deliverable, media, editorial, provenance. It can also be imported and used as a library. DRX color grading writes are calibrated in real time for Resolve Studio, and structured writes are verified by reading back through the panel. You can use npx davinci-resolve-mcp setup to install Node dependencies and register, or npx davinci-resolve-mcp sync to repair an existing installation.
Bradford Post Assistant (closed beta): The maintainer has also built Bradford Post Assistant on this open-source foundation, a local AI assistant desktop app for post-production, where footage does not leavethe workstation. It is currently in closed beta, and access can be requested at bradfordoperations.com/software/post-assistant. The open-source server itself is complete and can be used independently.
