AskEditor
Server Details
The agentic video editor for ChatGPT and Claude. Ask in your chat and it cuts, captions, adds music, titles and motion graphics on a real timeline you can open and adjust.
- Status
- Healthy
- Uptime
- 100.0% over the last 24 hours
- Last Tested
- Transport
- Streamable HTTP · MCP 2025-11-25
- URL
- ChatGPT
- –
- Claude
- –
TDQS
Scored across 48 tools
Most editing tools are carefully delineated (add_keyframe vs set_keyframes vs set_clip_animation, trim_clip vs split_clip), but there are real overlaps: set_canvas_aspect_ratio and set_export_settings both set the output frame, and get_timeline vs open_project vs watch_edit vs read_clip all read state. The meta-tools (call_tool, run_tools, edit_timeline, ask_askeditor) form a second layer whose boundaries an agent must reason about, adding friction to selection.
Almost everything follows snake_case verb_noun (add_text, set_clip_transform, list_media, get_export, delete_clip), which is highly predictable. Minor deviations: read_project alongside get_* readers, and the awkward ask_askeditor, plus meta names (call_tool, run_tools) that sit outside the verb_noun domain vocabulary.
48 directly exposed tools is far past a comfortable set, and the descriptions openly admit that many more (captions, transitions, ripple edits, audio, tracking) are reachable only through call_tool. The surface is heavy enough that an agent must triage, and several meta-tools exist largely to compensate for the breadth.
The core video-editing lifecycle is well covered: project/workspace management, media import and upload, timeline assembly, transforms, animation, keyframes, transitions, export, and motion graphics. Gaps are handled by indirection rather than absence — captions, ripple edits, asset deletion and bulk rendering exist only via call_tool/skills, so an agent that does not load a guide hits a dead end.
Publisher details
- Operator
- AskEditor · Publisher source
- Operator website
- https://askeditor.com
- Vendor relationship
- First-party
- Documentation
- https://askeditor.com/docs/mcp/connect
- Privacy policy
- Not provided
- Trust center
- Not available
- Restrictions
- A free AskEditor account connects, by OAuth or an API key. Generating images, video or voice uses credits from a paid plan or a credit pack. · Publisher source
Available Tools
48 toolsadd_keyframeAdd keyframeAInspect
[write] Add one keyframe (or merge into the one within a frame of atMs) on a visual clip. atMs is clip-local, and the other keyframes stay. Use to build or adjust an animation one point at a time. Not for replacing or clearing them all (use set_keyframes). Returns the changed clips' rows (with handles) and counts. Related: set_keyframes, get_timeline.
Inputs: project_id (required), x, y, id, +10 more.
add_markerAdd timeline markerAInspect
[write] Drop a ruler marker, optionally labelled. Skipped if one already sits within 120ms. Markers are notes for the person and change nothing in the video. Use to flag beats, cut points or spots to review. Not for captions or on-screen words (use add_caption_line or add_text). Returns the refreshed timeline table, markers (ms, label), tracks and transitions. Related: remove_marker, control_editor.
Inputs: project_id (required), ms, items, label.
add_textAdd text layerAInspect
[write] Add a text layer (title, headline, lower third, label) on its own top lane; style fields override the preset. items places every word of a sheet in one save. Use for on-screen words that are not spoken captions. Not for captions (use add_caption_line or import_captions) or changing existing text (use set_text). Returns the changed clips' rows (with handles) and counts. The new layer's handle is in changed. Related: set_text, set_clip_animation, align_clip.
Inputs: project_id (required), x, y, bg, +22 more.
add_timeline_clipAdd media clipAInspect
[write] Add a media clip from ONE of assetId or src, on the first free lane at startMs. items places a whole cut list in one save. Use to lay down footage, b-roll, images or music. Not for text, shapes or captions (use add_text, add_shape, add_caption_line), or to push later clips along (use insert_clip_ripple). Returns the changed clips' rows (with handles) and counts. The new clip's handle is in changed. Related: list_media, list_assets, trim_clip, set_clip_transform.
Inputs: project_id (required), src, kind, items, +8 more.
ask_askeditorAsk AskEditor's agentADestructiveInspect
[write · costs credits] Takes project_id and instruction. Hand an instruction to AskEditor's own agent, which applies its house skills and runs the tools itself (billed as a chat turn plus whatever it makes). Returns its answer, the tools it ran and, when it exported a film, export with its download_url (pass that on; never retype it). Use it for craft work (a hook, a camera, a title card, a caption style); use the direct tools for precise edits.
Inputs: project_id (required), instruction (required), answer, approve.
call_toolCall a toolADestructiveInspect
Run any AskEditor tool by name, including the many that are not in your tool list: the captions, motion graphics, animations, keyframes, transitions, audio, tracking, brand and board tools. Load the tools/ guide that owns it first (load_skill), then pass its arguments in args. Guards, approvals and results are exactly those of calling the tool directly.
Inputs: name (required), project_id (required), args.
complete_uploadFinish file uploadAInspect
[write] Assemble an upload's parts into a library file, measure it, and optionally place it on the timeline. Use after every part of a create_upload has been PUT; pass each part's ETag exactly as returned, and over MCP the project's project_id. place lays it like add_timeline_clip. Not for a web address (use import_media_from_url). Returns asset_id, kind, name, duration_ms, width, height, and placed {handle, lane, start_ms, end_ms} when placed. Related: create_upload, add_timeline_clip, list_media.
Inputs: parts (required), upload_id (required), project_id (required), place.
create_editor_projectCreate new projectAInspect
[write] A new project in your workspace, optionally opening on a library file (from list_media or an upload) as its first clip. Only when the person asks for a new project or has none. A project holds a board (script, scenes, generated images and video) and a timeline, so never make a second one for the same film: work in the project they opened or named. When you already have a project in hand it is refused unless person_asked is true. Returns project_id, title, kind, and its links as get_editor_url gives them.
Inputs: title (required), aspect_ratio, person_asked, workspace_id, +1 more.
create_uploadStart file uploadAInspect
[write] Open an upload of a local file (video, image or audio, up to 4 GB) and sign a URL for each part. Use for a file on your machine: PUT each part to its url, keep each ETag, then complete_upload. For parts past the first 50, or after 15 minutes, call again with upload_id and part_numbers. Not for a web address (use import_media_from_url). Returns upload_id, part_size, part_count, parts [{part_number, url}], how (a shell recipe) and next_part_numbers when more remain. Related: complete_upload, import_media_from_url, list_media. Full docs: load_skill with tools: [name].
Inputs: project_id (required), bytes, items, filename, +3 more.
delete_clipDelete clipsADestructiveInspect
[write] Remove clips, leaving a gap (ripple_delete_clip closes it). Use to drop a clip where the gap should stay. Not for closing the gap (use ripple_delete_clip or close_gap) or removing a motion graphic and every clip of it (use delete_asset). Returns the changed clips' rows (with handles) and counts. Related: ripple_delete_clip, close_gap, get_timeline. Full docs: load_skill with tools: [name].
Inputs: project_id (required), id, ids, items.
edit_timelineEdit timeline in one goADestructiveInspect
[write] Several timeline edits in one call, applied in order: each op is a timeline editing tool and its arguments (move_clip, trim_clip, split_clip, add_text, set_clip_transform, set_clip_animation, add_keyframe, add_marker…) or a caption edit (edit_subtitle_line, add_caption_line, position_subtitles, toggle_subtitles, set_subtitle_style). Use it instead of calling those tools one by one. Stops at the first edit that fails and says which. Returns what changed across all of them (changed clips, removed handles, and tracks/transitions/markers when they moved), plus a link to the whole timeline.
Inputs: ops (required), project_id (required).
export_videoExport the final videoAInspect
[write] Start the project's final export with set_export_settings' settings, on AskEditor's servers. Free includes 3 of those in all; after them the film is made in the person's editor beside the chat, on their computer, then uploaded (open it first). Use when the edit is finished and the person wants the file. Runs up to ~20 min; a call while one runs returns it. Refuses an empty timeline. Not for one graphic (use render_asset). Returns render_id, status, progress, watermarked and the settings used. Call get_export with wait_seconds 45 until status is done or failed. Related: set_export_settings, get_export, export_captions. Full docs: load_skill with tools: [name].
Inputs: project_id (required).
get_asset_metadataInspect clip media fileARead-onlyIdempotentInspect
[read] Inspect a clip's underlying media file: pixel width/height, aspect, fps, real duration, and whether it has an audio track. Use before placing or scaling B-roll so you know its shape. Not for a compound clip (use get_compound_children). Returns width, height, aspect, fps, hasAudio, fileDurationMs and clipDurationMs. Related: get_timeline, set_clip_transform, get_compound_children. Full docs: load_skill with tools: [name].
Inputs: project_id (required), id, items.
get_clip_transcriptGet clip transcriptAInspect
[read · costs credits] Read a clip's speech. Default 'vtt' is a compact WebVTT transcript ('00:12.340 --> 00:15.020 text' cues in TIMELINE time) — use it to understand what is said. Use to pick the lines for a hook, a title card or a reel, or to cut on a word ('words' gives per-word timings). Free within the plan's monthly minutes; kept on the file. Not for putting captions on screen (use generate_captions). Returns wordCount, startMs, endMs, credits_used and vtt, words or text. Related: add_hook, add_title_card, make_reel, generate_captions. Full docs: load_skill with tools: [name].
Inputs: project_id (required), id, toMs, items, +7 more.
get_editor_urlGet editor linkARead-onlyIdempotentInspect
[read] The project's links. url opens it in AskEditor: the one to hand the person. pane_url is the same editor with its own chat closed (null only for a talking film): if your app has a built-in browser pane, open it there so the person watches the edit beside this conversation. Pass pane: true when you are about to open that pane: the answer then carries browser_handoff.url, which opens it already signed in. That link works once, for two minutes; open it in the pane and never show it. A pane is only ever opened on a link from this tool, never on an address you wrote yourself.
Inputs: project_id (required), pane.
get_exportCheck the exportARead-onlyIdempotentInspect
[read] Read the project's latest export, or the one named: its status (queued, rendering, done, failed or stopped) and progress. Use after export_video, with wait_seconds 45, until status is done or failed. When done, hand over download_url: public and signed, it works for 3 days; call again for a fresh one. Not for starting an export (use export_video). Returns render_id, status, progress, watermarked; when done download_url, expires_at, size_bytes, duration_seconds, width, height; when failed error. Related: export_video, export_captions, get_social_state. Full docs: load_skill with tools: [name].
Inputs: project_id (required), items, render_id, wait_seconds.
get_timelineRead the timelineARead-onlyIdempotentInspect
[read] Read the timeline as a table: each clip's id handle, kind, label, lane, start/end/dur ms, trim, speed; 60 rows a read, or only fromMs–toMs, a kind or a lane. Use first, and again after the person may have edited, for the handles every edit tool takes. Not for what a clip shows or says (watch_clip, get_clip_transcript). Returns clipCount, totalDurationMs, timeline (the table), tracks, transitions, markers. Related: add_timeline_clip, move_clip, trim_clip.
Inputs: project_id (required), kind, lane, toMs, +2 more.
import_media_from_urlImport media from URLAInspect
[write] Download a public video, image or audio file into the workspace library (video up to 400 MB, audio 100 MB), measure it, and optionally place it on the timeline. Use for a direct link to an mp4/mov/webm, jpg/png/webp/gif or mp3/wav/m4a file. A private or local address is refused. Not for a web page (use find_site_assets) or a file on your machine (use create_upload). Returns asset_id, kind, name, duration_ms, width, height, reused, and placed {handle, lane, start_ms, end_ms} when placed. Related: list_media, add_timeline_clip, create_upload. Full docs: load_skill with tools: [name].
Inputs: project_id (required), url, kind, name, +2 more.
list_assetsList motion graphicsARead-onlyIdempotentInspect
[read] List the motion graphics this project can place — its own and the brand's — with status, format, length, and whether each has code yet. Use before rewriting, rendering, scoring or deleting a graphic. Not for photos, video or audio files (use list_media). Returns assets [{id, title, kind, format, status, hasCode, durationMs}], 100 a page, total, next_offset. Related: write_asset, render_asset, delete_asset.
Inputs: project_id (required), limit, offset, search.
list_mediaList media libraryARead-onlyIdempotentInspect
[read] List the workspace's media library (images, video and audio added or generated), newest first; filter by kind or name. Use to find a file to place. Not for motion graphics (use list_assets). Returns each item's id (for the TIMELINE: add_timeline_clip({ assetId: id })) and its timelineSrc, total and next_page. Related: add_timeline_clip, search_media, get_asset_metadata.
Inputs: project_id (required), kind, page, search.
list_projectsList projectsARead-onlyIdempotentInspect
[read] A workspace's projects, newest first (50). Without workspace_id, your default workspace. When the person names a project, pass search with words from its title: the list holds only the 50 newest, and search reaches the rest.
Inputs: search, workspace_id.
list_skillsList skillsARead-onlyIdempotentInspect
[read] AskEditor's skills, in two layers. how_the_tools_work: one tools/ guide per group of tools (tools/timeline, tools/animation, tools/captions, tools/sound, tools/motion-graphics, tools/tracking, tools/camera, tools/board…), each with the tools it owns; every tool is in one. how_to_make_things: guides for making one kind of thing well (hooks/guide, captions/guide, talking-video/guide…), each with the tools/ guides it needs. Load one with load_skill.
Inputs: no named parameters.
list_templatesList your templatesARead-onlyIdempotentInspect
[read] The workspace's own templates, newest first: each one's ref, name, line, its plan's latest version and the film it was made from. A template is a plan the person saved from one of their films, to make the next one the same way. Use it when the person names one of their templates; open_plan with its ref makes a new film from it. Only this workspace's: templates are never shared.
Inputs: workspace_id.
list_workspacesList workspacesCRead-onlyIdempotentInspect
[read] The workspaces you are in.
Inputs: no named parameters.
load_skillLoad a skillARead-onlyIdempotentInspect
[read] A skill's whole guide, and the tools it brings. A tools/ guide (e.g. tools/captions) comes with each of its tools' description and input schema, ready to call; a guide for making something (e.g. hooks/guide) names the tools/ guides it needs. Load a tools/ guide before you first use its tools. tools: [up to 10 names] gives those tools' full docs instead. Run a tool directly when it is in your tool list, any other with call_tool.
Inputs: name, tools, loaded.
move_clipMove clipAInspect
[write] Move a clip to a new start time and/or lane. Dropped on an occupied lane it takes the lane and what it meets goes up a lane; a same-lane collision bumps to the next free lane. Use to reposition one clip freely. Not for swapping with a neighbour (use move_clip_to_neighbor), lining up with another clip's edge (use snap_clip) or shifting everything after a point (use ripple_timeline). Returns the changed clips' rows (with handles) and counts. Related: get_timeline, snap_clip, reorder_clip.
Inputs: project_id (required), id, items, track, +1 more.
open_planOpen a getPlan.md planAInspect
[write] Takes plan (and optional project_id). Use a plan: one of the person's own templates (its ref from list_templates, remix/), or a plan from getPlan.md, a ready-made way to edit a video in a named style. plan is a ref, a getPlan.md slug (paper-cut-talking-head), its address, or the plan a kit comes with (documentaries/paper-short). With project_id the plan is put on that project, which is right when the person's footage is already in one; without it a new project is made for the plan. Returns the project and its links, the plan's kickoff (what it does, as a first message) and the skills to load before working. Use it when the person names a getPlan.md plan or a kit's plan, or pastes an address.
Inputs: plan (required), inputs, project_id, workspace_id.
open_projectOpen projectARead-onlyIdempotentInspect
[read] Start working on a project: its timeline (tracks, clips and their ids). Its board, where scenes, images and video are generated, is the same project: read it with get_video_assets_state. The project's chat in AskEditor shows that you are working on it. Call it first, and again to re-read the timeline after changes.
Inputs: project_id (required).
read_projectRead projectARead-onlyIdempotentInspect
[read] Read this project: its title, how long the video should be, what shape it is, who presents it and in what subtitle style, who it is for and how it should sound, and whether a script has been written yet. Use when starting on a talking project's script, to write to the right length and shape. Not for the board's scenes and assets (use get_video_assets_state). Returns title, target_seconds, target_words, aspect_ratio, presenter, subtitle_style, has_script, scenes and script_seconds. Related: read_script, write_script, set_length. Full docs: load_skill with tools: [name].
Inputs: project_id (required).
remove_markerRemove timeline markerADestructiveInspect
[write] Remove the ruler marker nearest to a time, within 120ms. With none that close it fails and changes nothing. Use to clear a flagged spot once it is dealt with. Not for clips (use delete_clip). Returns the refreshed timeline table, markers (ms, label), tracks and transitions. Related: add_marker, get_timeline.
Inputs: project_id (required), ms, items.
render_assetRender motion graphicAInspect
[generate · costs credits] Render one motion graphic to a file (an overlay as alpha WebM and a ProRes .mov, full-frame as MP4) with a still; costs credits, so only when asked. Use only when the person wants the file; the editor's preview does not need it. On failure, fix the code with write_asset and render again. Not for many at once (use render_all_assets). Returns asset (with status) and videoUrl. Related: write_asset, review_asset, render_all_assets. Full docs: load_skill with tools: [name].
Inputs: project_id (required), items, assetId, approval_id.
request_budgetRequest a budgetADestructiveInspect
[write · asks the person] Ask once for a budget of credits on one project, so a run's paid steps do not each wait for a yes. Ask for what the run will spend (a plan's quote) before its first paid step. A new budget replaces what is left of the last one on that project, it does not add to it, so ask again only for everything the run still has to spend. The person approves it like a paid call, and you call again with approval_id as for any approval. Once approved, calls on that project that cost credits run without asking while the budget covers them, for 12 hours. A call that deletes something still asks, and the day's limit and the balance still apply. Answers the budget, what is left, what it replaced and when it ends.
Inputs: credits (required), project_id (required), approval_id.
run_toolsRun tools at onceADestructiveInspect
Run several AskEditor tools in one request: up to 20 calls, 4 at a time (inOrder: true runs them one after another). Each call is that tool called on its own: its arguments, guards, approvals and credits apply to it alone, so a paid call answers needs_approval while the others run. Use it for independent steps: write several graphics, read several transcripts, look at several moments. Not for get_editor_url, load_skill or this server's other session tools. Returns one result per call, in order.
Inputs: calls (required), inOrder, project_id.
search_mediaSearch stock mediaAInspect
[read] Search stock providers for engaging real assets and (optionally) add the best hit to the library. Use for B-roll, a reaction GIF, a Lottie animation or a photo the project does not have; place an added one with add_timeline_clip. Not for the project's own files (use list_media). Returns up to 12 candidates {url, previewUrl, kind, title}, and with addTop the added {id, name, url}. Only configured providers work; one that is not answers with the sources that are. Related: list_media, add_timeline_clip.
Inputs: project_id (required), items, query, addTop, +1 more.
set_canvas_aspect_ratioSet canvas aspect ratioAInspect
[write] Set the PROJECT / EXPORT canvas aspect ratio — the frame the preview, every rendered asset and the composite export all use. Use whenever the deliverable's shape (16:9, 9:16 or 1:1) differs from the canvas, before making or rendering graphics, which render at the canvas size. It does not re-fit existing clips (use set_clip_transform). Answers with aspectRatio and the exact width and height to design against. Related: get_timeline, set_clip_transform, render_all_assets. Full docs: load_skill with tools: [name].
Inputs: project_id (required), aspectRatio (required).
set_clip_animationAnimate clipADestructiveInspect
[write] Animate a video, image, text or shape clip with entrance, exit and loop presets. Omitted sides are kept. "none" removes a side, and ids animates several. Use for quick motion on titles, stickers and stills. Not for custom moves (use set_keyframes) or cuts between clips (use set_transition). Returns the changed clips' rows (with handles) and counts. Related: set_keyframes, set_transition, add_text.
Inputs: project_id (required), id, in, ids, +6 more.
set_clip_timingSet clip timingAInspect
[write] Set a clip's start and/or its timeline length directly. It keeps the source in-point (300 ms minimum, capped by the source unless it loops). Use to hold an image or a title longer, or to place a clip at an exact time. Not for changing where in the source it starts (use trim_clip). Returns the changed clips' rows (with handles) and counts. Related: trim_clip, set_clip_speed, ripple_timeline. Full docs: load_skill with tools: [name].
Inputs: project_id (required), id, items, startMs, +1 more.
set_clip_transformTransform clipAInspect
[write] Position, scale, rotate, crop or fit a visual clip; only passed fields change. ids applies one transform to several clips. Use to reframe after a new aspect ratio, make a picture-in-picture or punch in. Not for audio or compound clips, snapping to a frame edge (use align_clip) or movement over time (use set_keyframes). Returns the changed clips' rows (with handles) and counts. Related: align_clip, set_keyframes, set_export_settings.
Inputs: project_id (required), x, y, id, +9 more.
set_export_settingsSet export settingsAInspect
[write] Set the frame and the export settings. The frame is 9:16, 16:9 or 1:1 and every clip is fitted into it; the export takes format, frame rate, resolution, quality, background colour and loudness. Only passed fields change, and the frame shows at once. Use before the person exports, or when the deliverable's shape changes, then reframe clips with set_clip_transform. Not for rendering a motion graphic (use render_asset). Returns ok, aspectRatio, format, fps, resolution, quality, backgroundColor, loudness. Related: set_clip_transform, align_clip, set_canvas_aspect_ratio. Full docs: load_skill with tools: [name].
Inputs: project_id (required), fps, format, quality, +4 more.
set_keyframesSet clip keyframesADestructiveInspect
[write] Replace a clip's property keyframes. Values ease between keyframes and hold outside them. Keyframes animate position, scale, rotation, opacity, blur and colour at clip-local times; an empty list clears them. Use for push-ins, custom moves, blur-ins and fades to black and white. Not for adding one point to what is there (use add_keyframe) or preset motion (use set_clip_animation). Returns the changed clips' rows (with handles) and counts. Related: add_keyframe, set_clip_transform, set_clip_animation. Full docs: load_skill with tools: [name].
Inputs: project_id (required), id, items, keyframes.
set_textEdit text layerADestructiveInspect
[write] Change a text clip's words and/or style. preset replaces the style first; passed fields merge over it. ids restyles several at once. Use to fix wording or restyle titles. Not for caption lines (use edit_subtitle_line), the caption style (use set_subtitle_style) or placement (use set_clip_transform). Returns the changed clips' rows (with handles) and counts. Related: add_text, set_clip_transform, set_clip_animation.
Inputs: project_id (required), bg, id, ids, +20 more.
set_transitionSet transitionADestructiveInspect
[write] Add, change or remove the transition where two clips meet on a lane. It belongs to the incoming clip. It can set every such seam at once. Use for fades, wipes, slides and zooms between shots; the clips must meet end to end first (use close_gap or move_clip). Not for audio or captions, or a clip's own entrance (use set_clip_animation). Returns the refreshed timeline table, transitions (into: the incoming clip's handle, type, durationMs), tracks and markers. Related: close_gap, get_timeline, set_clip_animation.
Inputs: project_id (required), id, all, type, +2 more.
split_clipSplit clipAInspect
[write] Split a clip at absolute timeline time(s), N cuts in one call. Each cut sits at least 300ms from the clip's ends and its neighbours. For a compound's child, times are compound-local. Use before removing, moving or restyling part of a clip. Not for trimming an end (use trim_clip). Returns the changed clips' rows (with handles) and counts. Each new piece's handle is in changed. Related: get_timeline, ripple_delete_clip, delete_clip.
Inputs: project_id (required), id, atMs, items.
trim_clipTrim clipAInspect
[write] Trim a clip's source window. Clamped to the media and a 300ms minimum. It sets the source in-point and/or the length on the timeline, and optionally the start. Use to cut the head or tail off a take. Not for cutting a clip in two (use split_clip) or closing the gap a trim leaves (use close_gap or ripple_timeline). Returns the changed clips' rows (with handles) and counts. Related: get_timeline, split_clip, close_gap.
Inputs: project_id (required), id, items, startMs, +2 more.
wait_for_creditsWait for creditsARead-onlyIdempotentInspect
[read] After a tool answered waiting_for_credits: holds until the person has added the credits that call needs, up to ten minutes, then answers covered: true. Call it right after telling the person, in one plain sentence, how many more credits are needed and where to add them (the waiting_for_credits answer says where). When it answers covered, call the tool again with the same arguments. When it answers covered: false, do as its message says and do not retry the tool. Never name a tool to the person.
Inputs: hold_id (required), timeout_seconds.
watch_clipWatch a clipARead-onlyIdempotentInspect
[read] WATCH footage as timestamp grids: a timeline clip (clipId, timeline ms) or a library video (assetId, file ms), nothing laid over it. To see what happens and when: find a moment, pick cut points, see where a person stands. The finished edit is watch_edit; exact positions read_clip. Sheets [{asset_id, grid, fromMs, toMs, cells}]; continueFromMs when more remains. Related: watch_edit, read_clip, get_clip_frames, detect_scenes. Full docs: load_skill with tools: [name].
Inputs: project_id (required), fps, toMs, items, +7 more.
watch_editWatch the editARead-onlyIdempotentInspect
[read] WATCH THE EDIT as it will export, as timestamp grids, overlays and captions included. After every visible change, before saying it is done. Footage alone is watch_clip; exact positions read_clip. Sheets [{asset_id, grid, fromMs, toMs, cells}]; continueFromMs; layout {errors, warnings, findings [{severity, rule, message, fix, atMs}]} in pixels; failed_graphics [{asset_id, title, error}] left out. Fix every error as its fix says. Related: watch_clip, read_clip, get_timeline. Full docs: load_skill with tools: [name].
Inputs: project_id (required), fps, toMs, fromMs, +4 more.
write_assetWrite motion graphicADestructiveInspect
[write · costs credits] Make a motion graphic, or rewrite one. Written as a React TSX module; the component gets the brand kit as a theme and, over footage, where faces, objects and burnt-in text are, and a new one is placed on the timeline. Use for lower-thirds, callouts, charts, kinetic type and stickers made to order; read get_motion_doc before non-trivial code. The preview is live, nothing is rendered. Returns asset {id, status}, its placement and the clip's {handle, lane}. Related: get_motion_doc, render_asset, list_assets, set_asset_sound. Full docs: load_skill with tools: [name].
Inputs: project_id (required), kit, bake, code, +22 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
Edit your own videos by chatting in Claude or ChatGPT: cut, caption, add music and export.
Agentic AI video editor: edit a real timeline, generate clips, music and voice, then render.
Edit video by talking to your AI — search footage, cut timelines, apply effects, add captions.
AI video editor your agent can drive: plan scenes, edit a multi-track timeline, and render MP4s.
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceEnables AI agents to build and edit a real video timeline — cuts, text, shapes, keyframes, captions and music — in a live studio the user can watch and edit alongside, rendering actual frames so the agent can verify its own work instead of guessing. Finishes by exporting an MP4 or carrying the timeline into DaVinci Resolve, Premiere Pro, CapCut or Remotion.2MIT
- FlicenseNot gradedqualityAmaintenanceEnables users to turn raw video clips into finished, beat-synced montage cuts through natural-language chat, automating shot selection, music matching, subtitles, transitions, picture-in-picture, and rendering. Supports Claude Code, Codex, Hermes Agent, and OpenCode with previewable timelines before rendering.-
- FlicenseNot gradedqualityAmaintenanceEnables users to edit raw video clips through natural-language chat in AI coding agents, turning footage into finished montage videos with beat-synced cuts, music matching, subtitles, transitions, and picture-in-picture.-
- FlicenseNot gradedqualityBmaintenanceEnables AI agents to autonomously edit raw video footage into publish-ready videos with millisecond-accurate cuts, Whisper transcription, karaoke captions, motion graphics, and high-CTR thumbnails.6-
Glama MCP Gateway
Add one secure layer between your agents and this server.