# Local Control Panel The DaVinci Resolve MCP ships with a local, single-user browser control panel for inspecting Resolve state, running source-safe media analysis, drilling into analyzed clips and shots, fixing analysis output inline, reviewing timeline edit history, driving Resolve's local AI operations, and managing preferences. The panel is a local HTTP server (default `http://127.0.0.1:8765`). It does not expose a network listener beyond loopback and does not modify source media. ## Launching the panel ```bash # From source venv/bin/python -m src.control_panel # From the npm package npx davinci-resolve-mcp control-panel # From a chat client via MCP resolve_control(action="open_control_panel") ``` Once running, `resolve_control(action="control_panel_status")` checks the pidfile and `resolve_control(action="close_control_panel")` stops it. ## Navigation and deep links Top-level sections: **Overview**, **Media** (Inventory / Review / History / Edit Plans), **AI Console**, **Setup** (Resolve / MCP / Storage / Tools / Media Pool History), **Docs**, and **Preferences** (Analysis / Caps + Safety / Metadata and Markers / Paths and Workflow / MCP Updates). The project selector on the right scopes the panel to an analysis context; **View All Projects** opens a read-only browser over the Resolve project database with confirm-gated loading. Every view is addressable by URL hash, so links can be bookmarked or pasted into chat: ``` #overview #aiconsole #analysis/media #analysis/review #analysis/review/history #analysis/review/clip/ #analysis/review/clip//shot/ #analysis/review/clip//transcript #analysis/review/plans #analysis/review/plans/ #diagnostics/mcp #preferences/caps ``` ## Tour ### Overview ![Control panel overview](../images/control-panel/01-overview.png) The at-a-glance summary for the active project: Resolve connection state, source clip counts, analysis progress, search-index status, and the source-media safety posture. When the project has no source clips yet, the view shows a get-started card with a suggested chat prompt and a **Copy prompt** button — the panel is summoned by conversation, and its empty states point back into it. ### Media → Inventory ![Inventory view](../images/control-panel/06-inventory.png) Inventory walks the Resolve Media Pool read-only and filters to source clips so timelines, compounds, titles, and generated items stay out of analysis queues. Filters cover bin, media type, clip status, and analysis status, with optional auto-refresh polling. From here you can select analyzable clips, copy a ready-made analysis prompt for your chat session, or launch a supported chat client directly. ### Media → Review ![Review bin grid](../images/control-panel/02-review-bin-grid.png) Review is the browser for analyzed clips. The readiness strip summarizes the evidence base (analyzed / superseded / vision-pending / warnings) with humanized source-trust and analysis-layer chips. Cards show a representative thumbnail, duration, shot count, status, and summary one-liner; grid and list layouts are available, and full-text search covers clips, summaries, tags, and transcripts once the search index is built. When a local text-embedding backend is detected (ollama with `nomic-embed-text`, or sentence-transformers), a `Semantic` toggle appears next to the search box — it searches by meaning instead of exact words, ranking clips, shots, and transcript lines by similarity to your query. Once cross-clip entity detection has run (`detect_entities` + a one-frame-per-cluster confirmation in chat), a `Recurring across this bin` card lists the labeled people, places, and objects with their shot counts. ### Clip detail ![Clip detail with shot strip](../images/control-panel/03-clip-detail.png) The clip view shows the full summary, tags, star rating, editorial notes, and a contact sheet of every detected shot. From here you can open the clip in Resolve, jump to the transcript, or click into any shot. `Deepen analysis` copies a ready-made chat prompt that asks your MCP session to run the opt-in deep shot pass (cost estimate first, then per-shot Visual / Content / Editorial / Cuttability fields). ### Transcript ![Transcript view with word-level timing](../images/control-panel/04-transcript.png) Each segment shows its start time, the cleaned sentence, and the original machine transcription beneath. Inline edits are preserved across re-analysis, and `Re-transcribe with words` re-runs the configured local transcription backend (Whisper by default) without re-doing visual analysis. ### Shot detail with inline correction ![Shot detail with fields and frames](../images/control-panel/05-shot-detail.png) The shot view shows the analysis fields for the shot alongside the frames the vision pass actually sampled, each labeled with why it was chosen ("Shot start", "After cut", "Flash frame", "Motion peak"). Subjective fields are editable inline; edits are kept with the clip's analysis and merged on top of future re-analysis so human notes survive fresh vision runs. `Open in Resolve` jumps straight to the clip in the source viewer with the shot's mark in/out set. The field groups (Visual, Content, Production, Editorial, Cuttability) are filled by the opt-in deep vision pass — `Deepen this shot` copies a chat prompt that runs it for just this shot, estimate first. The Relationships group (same_setup_as / continues_from / alt_take_of) fills from the cross-shot relationships pass (`detect_shot_relationships` → vision-confirm → `commit_shot_relationships`); `continues_from` is shown on the continuing shot. ### Media → History ![Timeline history](../images/control-panel/10-history.png) The timeline edit history: archived timeline versions and brain edits per timeline. Begin/end labeled runs to group related edits, archive the current timeline before risky work, and select any timeline to inspect its version chain and the edits between versions. Every destructive timeline edit made through the MCP archives a version here automatically. ### Media → Edit Plans ![Edit plan browser](../images/control-panel/11-edit-plans.png) Dry-run plans saved by the edit engine (`edit_engine` actions `plan_selects` / `plan_tighten` / `plan_swap`). The list shows each plan's kind, summary, save time, and an `executed` chip once it has run; plans that fail their fingerprint check appear as a warning row instead of being hidden. Opening a plan shows its full evidence: selects decisions render with shot thumbnails, rank, duration, and rationale (with a deep link to each shot page); tighten plans list every dead-air lift with its transcript-gap evidence; swap plans show the current item and each numbered alternate with its similarity score. The panel never executes a plan. Each detail view carries a copyable chat prompt (per kind, with the `alternate_index` placeholder for swaps) — paste it into your MCP chat session, which holds the confirm-token gate and the versioned execution path. Executed plans keep their execution readback inline. ### AI Console ![AI Console with governance](../images/control-panel/09-ai-console.png) Runs Resolve's local AI operations (audio classification, IntelliSearch, slate analysis, transcription, motion deblur, speech generation) on the current Media Pool folder or a specific clip. Capability dots show what this Resolve build exposes and which AI Extras are required. The **Governance** card sets the per-session tier (off / lenient / standard / strict) for the two media-creating ops and the governance **mode**: *Advisory* warns in the confirm dialog; *Enforce* refuses an over-tier run until you raise the tier, relax the mode, or consciously override. Usage gauges track runs, media created, and render time against the active tier, and the AI-ops ledger (Preferences → Caps + Safety) records every run with the acting instance. ### Setup ![Setup / Resolve diagnostic](../images/control-panel/07-diagnostics-resolve.png) - **Resolve** — live connection, product/version, active page, project identity, and Media Pool inventory. - **MCP** — server identity, detected Resolve scripting paths, transport status (with a start button for the networked mode), and per-client install status with one-click install/repair for every supported harness. - **Storage** — the analysis root, search index, and jobs database paths and sizes for the active project. - **Tools** — runtime helpers (ffprobe, ffmpeg, Whisper backends, OpenCV) with ready/missing status and copyable install commands. - **Media Pool History** — the provenance log for destructive media-pool operations (deletes, replaces, relinks), kept separate from timeline brain edits. ### Preferences ![Preferences / Analysis defaults](../images/control-panel/08-preferences-analysis.png) Server-wide defaults for this MCP install (dashboard-only conveniences stay in the browser): - **Analysis** — vision/transcription defaults, slate detection, source-trust grading, depth, frame-sampling mode, summary style, report format. - **Caps + Safety** — token and frame budget presets (minimal / standard / generous / unlimited) with live usage gauges, per-clip/job/day caps, the Resolve AI-ops ledger, and destructive-edit safety rails. - **Metadata and Markers** — Resolve metadata write-back fields, timed-marker defaults, marker types and colors, overwrite policy. - **Paths and Workflow** — preferred analysis root, generated-media folder, post-operation page behavior, and a read-only map of where files live. - **MCP Updates** — update policy (prompt / notify / auto / never), release channel (stable / beta / dev), check cadence, apply/rollback with update history. ## Chat ↔ panel state sharing The panel and chat share focus state through `panel_state.json` (under the active project's analysis root). MCP actions: - `media_analysis(action="get_panel_state")` — read current focus (project, clip, shot) - `media_analysis(action="set_panel_state", params={...})` — set focus from chat - `media_analysis(action="session_start_context")` — bootstrap a chat session with the panel's current focus The dashboard polls `/api/panel_state` every 2 seconds while it has focus, backs off to roughly every 10 seconds when the window is unfocused, and stops polling entirely while the tab is hidden. ## Open in Resolve The shot view and clip view both have an Open-in-Resolve action. Under the hood this calls `media_pool_item(action="open_in_viewer", params={clip_id, mark_in, mark_out})`, which selects the clip on the Media page, loads it into the source viewer, sets mark in/out, and brings Resolve to the foreground via an OS-level window activation (AppleScript on macOS, PowerShell `AppActivate` on Windows, wmctrl/xdotool on Linux).