Best for
- Creators and developers who want AI to participate in editing while keeping professional control.
- Users who need inspectable, undoable editing with local data storage.
O
OpenChatCut
Community
Open-source, local-first, agent-native video editor
- Category
- Other #20 of 230
- Stars
- ★ 2.3k Very popular
- Transport
- Streamable HTTP
- Runtime
- Node.js 24.x
- Credentials
- Optional API key
- License
- AGPL-3.0
- Last commit
- Tools
- 3
68FMRS · C
OpenChatCut is an innovative open-source AI video editor that merges a professional timeline with agent-driven editing, offering reversible and inspectable workflows. Its MCP integration allows external agents like Codex and Claude Code to edit real projects over Streamable HTTP. While early in development, it is feature-rich and ideal for users who want AI assistance without sacrificing control.
Strongest · Documentation 16/20 Weakest · Reliability 8/20
Why each score
- Reliability 8/20
- Static review based only on the README and repo metadata. The README describes a plausible happy path: a Streamable HTTP MCP endpoint and an edit-session workflow (begin_edit_session / review_edit_session / get_edit_session) with draft isolation and atomic commits. However, the server manifest is not cached, so the declared tool list and input schemas cannot be matched against real behavior; no CI workflows or committed tests covering the MCP server's key paths were supplied, and edge-case/error handling is thinly documented. Static calibration caps reliability at 12 without verifiable execution evidence, and missing evidence is a deduction, so 8 (rounded down when unsure).
- Security and permissions 15/20
- The disclosed security posture is strong on paper: API keys remain server-side and must never reach the browser through VITE_ prefixed variables; external exposure requires a Bearer token (OPENCHATCUT_MCP_TOKEN); local binding and single-user scope are the defaults; sessions expose only draft-safe tools, excluding generation/export/deletion side effects; manual approval inside the editor is the default while auto mode is an explicit opt-out of review; LLM output, plugin packages and shaders are validated or sandboxed at trust boundaries; data-flow (local ~/.openchatcut store, configurable media directory, R2 credentials kept out of project JSON) is clearly disclosed; install examples contain no real secrets. No red-line issues were found. Deductions: none of this is verifiable from the provided source files, and auto mode intentionally bypasses project confirmation, so 15 rather than a higher score.
- Maintenance 15/20
- Evidence of sustained maintenance is present: the repo is not archived, the AGPL-3.0 license is explicit, the README references a bilingual CHANGELOG, GitHub Releases with installers for macOS/Windows/Linux, contributing guidelines, Discord and WeChat communities, an active-development status, and build/test/lint verification commands; ownership is clear. Deductions: there is no direct evidence of commit cadence, release frequency, dependency-update hygiene, or issue response times, and no security/vulnerability reporting channel (e.g., SECURITY.md) is mentioned, so full marks are not justified.
- Documentation 16/20
- The README is layered and detailed: quick start (installers and from source), built-in Agent authentication (API keys, Codex CLI sign-in, Claude via local MCP), the MCP endpoint URL, the edit-session workflow with tool semantics, approval modes, security model, data/privacy notes, environment variables, plus concrete client config examples for Codex (TOML) and Claude Code; limitations such as unsigned builds, early status, and the single-user bridge are disclosed. Deductions: the MCP manifest and tool-level schemas are absent, the 15 skills are not enumerated with parameters, and there is no dedicated MCP troubleshooting or cost/limits section.
- Setup experience 14/20
- The path from runtime to a working connection is clear and short: Node 24.x, npm install, copy .env.example, npm run dev, then either npx skills add 0xsline/OpenChatCut or manual client configuration (Codex TOML / Claude Code command); the repository root .mcp.json already carries the local connection, and desktop installers cover the three major platforms. Deductions: per static calibration, setup cannot exceed 15 without verifiable CI/test execution evidence, which is absent here; the MCP server is not standalone — the full editor must be running, and timeline tools additionally require an open project — and Node 24.x is a hard version requirement, so 14.
Static review · not runListed 2026-08-07
Read the FMRS scoring method →Fit and risk
What it can accessWrites / deletes local filesUses the network
Not for
- Teams requiring full cloud services and multi-tenancy.
- Users seeking one-shot generation of immutable videos.
Required permissions
- Local file system access (to store projects and media).
- API keys for external AI models and media services if configured.
- MCP endpoint binds locally by default; external access requires Bearer Token.
Risks and side effects
- Security risks if MCP endpoint is exposed without token, allowing unauthorized access.
- License risks from third-party dependencies.
- AI-generated content may be uncontrolled; review needed.
Setup
Before you start
Runtime：Node.js 24.x
OPENCHATCUT_MCP_TOKEN
Bearer token you set to protect an externally exposed MCP endpoint; clients must send Authorization: Bearer <token>.
Other optional settings (5)
OPENCHATCUT_EDITOR_URL
Public editor URL when self-hosting the MCP bridge, e.g. https://your-editor.example.com.
OPENCHATCUT_RENDER_CONCURRENCY
Tunes render concurrency; see .env.example.
OPENCHATCUT_MAX_ACTIVE_EXPORTS
Limits the number of concurrent heavy exports; see .env.example.
OPENCHATCUT_DISABLE_HARDWARE_ENCODING
Set to disable hardware encoding (VideoToolbox/NVENC) and fall back to software encoding.
OPENCHATCUT_H264_ENCODER
Overrides FFmpeg-side H.264 encoder selection; see .env.example.
- Install OpenChatCut desktop app or run from source (Node.js 24.x).\n2. Run npm install and copy .env.example to .env.local, configuring API keys for models/media services as needed.\n3. Start the editor via desktop app or local service (default http://localhost:5199).\n4. For Codex, add the MCP server to your Codex config.\n5. For Claude Code, run: claude mcp add --transport http openchatcut http://localhost:5199/api/external-mcp/mcp.
Check that it works
Run OpenChatCut (npm run dev, open http://localhost:5199) and configure the MCP endpoint http://localhost:5199/api/external-mcp/mcp; then confirm begin_edit_session, get_edit_session, and review_edit_session appear in the Codex/Claude Code tool list, or ask the agent to begin an edit session and return an editSessionId.
Troubleshooting
- Ensure Node.js version is 24.x and run npm install.
- Configure API keys properly in .env.local; missing keys are reported explicitly.
- If MCP connection fails, verify the editor is running and the project is open.
Things to try
Once connected, you can ask your AI assistant things like:
- Start an OpenChatCut edit session. Read the draft, add a scratch sound effect at 8 seconds on the second audio track, and add a glitch transition between the adjacent video clips, then submit the draft for review.
- Use the editSessionId to check the current edit session status and report whether the proposal was applied or rejected.
- Review the edit session in auto mode so the whole draft is applied to the timeline immediately.
Tools 3
begin_edit_session writes
Starts an edit session and returns an editSessionId to pass to all project read/edit tools.
get_edit_session read-only
Polls the edit session status (applied, rejected, or discarded).
review_edit_session writes
Submits the draft for review; in manual mode requires user approval, in auto mode applies immediately.
No matching tools
Use cases
Edit videos with natural language: describe the goal, the agent reads the project and produces verifiable edits.
Automate post-production: generate edits, transitions, captions, soundtracks, etc.
Integrate with Codex or Claude Code to directly manipulate projects via MCP.
Supported clients
Codex
Claude Code
Listed from the project's documentation, not tested by this site.
Overview
OpenChatCut is an open-source, local-first conversational AI video editor with a professional multi-track timeline, Agent Skills, MCP integration, and Remotion rendering. It lets Codex, Claude Code, and the built-in agent read, edit, and export real video projects that remain fully editable. Every edit is written to real tracks, clips, transitions, captions, effects, and media inside the project, so you can continue editing manually, undo/redo, save versions, or hand off to another agent.
Similar servers
Hevy MCP
77 · B
Manage your Hevy workout data from AI assistants via MCP.
Firecrawl MCP Server
75 · B
Firecrawl's official MCP server for web search, scraping, and structured extraction for AI agents
BioMCP
75 · B
One binary. One grammar. Evidence from the biomedical sources you already trust.
Compartment
75 · B
Encrypted, fully offline long-term memory for AI agents
Source revision 2e6f4a29b784 Data synced 2026-10-12 Read the FMRS scoring method