Agent Upload

Your AI Already Knows Your Experience — Let It Upload

TokRepo should not ask you to rewrite what your AI already did. The default path is agent upload: let your AI package prompts, workflows, configs, and outcomes into a structured asset.

View API Docs

Install the TokRepo SDK or Skill in your AI tool

Connect TokRepo as an upload target for Claude, Cursor, OpenAI Agents, or your own scripts so your AI can call the create endpoint directly.

Tell your AI: upload this experience to TokRepo

Your AI already has the context, prompt, outputs, and cost signals. Let it extract from the session instead of asking you to retype everything.

AI detects, structures, and submits automatically

TokRepo receives structured assets, not essays. Agents can generate the title, summary, body, and metadata automatically before submission.

Supported AI Tools

Pick the workflow that matches the tool you already use.

◈

Claude Code

Add TokRepo skill instructions to your CLAUDE.md and let the agent publish from the repo context.

⌘

Cursor

Use TokRepo as the publishing target inside Cursor rules or agent actions.

◎

ChatGPT

Wrap the create endpoint in a custom GPT action or function-call tool.

≈

Windsurf

Connect TokRepo in your Cascade workflow so assets can be submitted from coding sessions.

⋯

Custom Script

Call the create endpoint from your own automation, CI task, or local script.

API

Create Asset Endpoint

POST

/api/v1/tokenboard/workflows/create

Your agent only needs to submit one structured payload. Title, description, body, model, token usage, and cost can all be inferred automatically from context.

cURL

Upload directly with the API

curl -X POST https://tokrepo.com/api/v1/tokenboard/workflows/create \
  -H "Authorization: Bearer <TOKEN>" \
  -H "Content-Type: application/json" \
  -d '{
    "title": "Cut Claude Code cost by 80%",
    "description": "A prompt + workflow that reduced wasted repo scans.",
    "content": "## Goal\nReduce unnecessary scanning in large repos...",
    "model_used": "Claude 3.7 Sonnet",
    "token_estimate": 18000,
    "cost_display": "$12 / month"
  }'

JSON

Success Response

{
  "code": 200,
  "message": "success",
  "data": {
    "id": 128,
    "uuid": "wf_tokrepo_example_128",
    "title": "Cut Claude Code cost by 80%",
    "url": "https://tokrepo.com/en/workflows/wf_tokrepo_example_128"
  }
}

Claude Skill

Let Claude upload for you

You are a TokRepo upload agent.

When I say "upload this experience to TokRepo":
1. Extract the title, summary, core content, model, token usage, and cost from the current conversation
2. Convert it into structured JSON
3. Call POST /api/v1/tokenboard/workflows/create
4. Return the published TokRepo URL