Built for developer marketing teams

PandaStudio for developer marketing

Ship feature-launch videos, dev-tool demos, and changelog content with AI agents and a CLI you can pipe into CI.

The problem

You're on a developer-marketing or DevRel team at a dev-tools company. You ship feature videos, conference recap reels, weekly changelogs, and tutorial content. Your editing tool needs to fit how engineering teams work — scriptable, version-controlled, command-line-driven — not a SaaS dashboard with a per-seat license. Most tools fail this bar; the ones that don't (DaVinci) require a trained editor.

How PandaStudio fits

PandaStudio is built with a CLI + MCP server as a first-class surface. Every UI feature is also a verb-noun command. A GitHub Action can trigger a release-note video on every tag. Claude or Cursor can edit a recording from a single prompt. Project files are JSON in your repo. The whole workflow fits how DevTool teams already work — code-as-config, agent-driven, no per-seat cloud SaaS.

A typical workflow

1
1. Record (or hand a recording to the agent)
Use the desktop recorder, or hand any MP4 to the CLI: `pandastudio project.new --withMedia=...`
2
2. Script the edit
`pandastudio transcript.transcribe --id=$ID && pandastudio transcript.remove-fillers --id=$ID` — every command is shell-pipelinable.
3
3. Or hand it to Claude
From Claude Code: 'open the latest recording, remove fillers, add a release-note title card with v1.24, export landscape + vertical'. The agent calls every PandaStudio verb as a tool.
4
4. CI-trigger the release video
On `git tag v*`, run a workflow that records (or imports), edits, and publishes — same way you'd auto-deploy docs.
5
5. Repurpose
Same project exports landscape (YouTube), vertical (LinkedIn / X), and a square loop for embed-on-landing-page.

Features that matter most

Full CLI: 60+ verb-noun commands, every UI feature scriptable
MCP server: Claude / Cursor / Continue / Cline drive the editor as a tool
Project files are JSON — commit to git, diff changes, review like code
Local-first: dev-tool demos with internal infra never touch a SaaS server
Programmatic motion-graphics: HTML+GSAP escape hatch for branded title cards
One-time license, no per-seat cloud — fits internal-tool budget approval
Bundled @writepanda/cli (npm install -g) + @writepanda/mcp (any MCP client)
Built-in Claude Skill (npx skills add kamskans/pandastudio-skills) — agents auto-discover the workflow

Frequently asked questions

Can I run PandaStudio in CI?

The desktop app must be running on a Mac or Windows host (the Tier-3 PixiJS exporter uses WebGL/WebCodecs which need a real GPU and display). For headless server use, the agent-export dispatcher spawns a hidden BrowserWindow on the same machine to run exports, so CI runners with macOS or Windows hosts work — Linux pure-headless doesn't yet.

Is project data version-controllable?

Yes. .pandastudio project files are JSON — commit to git, diff changes, review like code. The schema is documented (see SKILL.md) and we use revision numbers so concurrent edits trigger conflict detection rather than silent overwrites.

Can I extend PandaStudio with custom verbs?

Right now the verb registry is internal — we ship new verbs in releases. The MCP escape hatch (`pandastudio_call`) covers any verb that's added after the MCP server version installed in your client. For team-internal extensions, the project-file format is open enough that a sidecar tool can manipulate projects directly.

How do I integrate with Claude Code or Cursor?

Two options. Claude users: install the bundled Claude Skill via Settings → Local automation → Install Skill (writes to ~/.claude/skills/pandastudio/). Cursor / Continue / Cline / other MCP clients: install @writepanda/mcp, point to it in your client's MCP config, and the 30+ static tools become available immediately.