API-first for agents
Give agents a real YouTube-to-text workflow today, then layer SDKs and MCP on top.
YT2Text is already usable from agents over the public REST API. The live path today is account → plan → API key → skill.md → async processing. This page is the human-facing entry point for that setup.
Quick start
Read https://yt2text.cc/skill.md Use https://api.yt2text.cc/api/v1 Auth: Authorization: Bearer sk_... Plan: Plus or Pro for API access Minimum plan for agent API usage: Plus
Batch, webhooks, PDF export, custom prompts: Pro
Internal roadmap is tracked in docs/project/AGENTS_NEXT_STEPS.md
How agent access works
The live integration path is intentionally simple: use the public API now, document the contract clearly, and avoid waiting for SDK or MCP surface area before agents can start shipping.
Step
Provision access
Create an account, choose Plus or Pro, and generate an API key in the dashboard. API access starts at Plus.
Step
Load the public skill
Point the agent to skill.md for the exact auth model, request flow, limits, and plan gates before it makes calls.
Step
Run async jobs
Submit a video, poll status, fetch results, and upgrade to Pro when you need batch jobs, webhooks, or PDF export.
Compatibility
Built for agents that can fetch Markdown and call HTTPS APIs
This is the boring compatibility layer you want first. If an agent runtime can read `skill.md`, keep secrets, and make authenticated HTTP calls, it can use YT2Text today.
Codex
GPT-based agents
Claude
Claude Code
OpenClaw
Hermes agents
custom internal agent runtimes
Operational instructions
The exact workflow an agent should follow
This mirrors the public skill and keeps the first version API-native. No separate package is required to get started.
- 1
Read
https://yt2text.cc/skill.md - 2
Store
YT2TEXT_API_KEY server-side - 3
POST
/api/v1/videos/process - 4
Poll
/api/v1/videos/status/{job_id} - 5
Fetch
/api/v1/videos/result/{job_id} - 6
Upgrade
Use Pro for batch, webhooks, and PDF export
Current platform status
What exists now vs. what is planned
| Capability | Status | Notes |
|---|---|---|
| Public REST API | Live now | Single-video async processing with transcript and summary payloads. |
| Agent skill file | Live now | Public `skill.md` and `skill.json` on the root domain for agent onboarding. |
| API keys | Live now | Created in the dashboard. Minimum plan: Plus. |
| Batch processing | Pro | Available today via `/api/v1/batch/*` for high-volume workflows. |
| Webhooks | Pro | Single-job completion callbacks are live. Batch completion delivery still needs product work. |
| First-party SDKs | Planned | JavaScript/TypeScript and Python first, then Swift and Kotlin. |
| MCP server | Planned | Small tool surface for submit, status, result, and usage checks after API/SDK hardening. |
Roadmap
The next steps after API-first launch
We should not block agent adoption on the perfect package story. The sequence is API now, SDKs next, MCP after the contract stabilizes.
Phase 1
API-first agent onboarding
Ship a clear human-facing page, public skill files, and docs so agents can integrate today without waiting for SDKs.
Phase 2
Official SDKs
Wrap auth, polling, retries, and typed responses into first-party packages so agents stop rewriting the same client logic.
Phase 3
MCP surface
Expose a narrow MCP layer once the API and SDK behavior are stable enough to support durable tool contracts.
Start now
Agents can integrate today over the public API.
Create an account, generate a key, hand the agent the public skill, and keep the first version boring and reliable.