What It Is
Godle's API v3 is a static JSON layer served directly from Vercel's CDN. Because it's static files rather than a server, there are no rate limits, no authentication headers, no CORS issues, and no downtime from server failures. You can fetch it from a browser, a Node script, a Python notebook, or a Cloudflare Worker.
The dataset covers 185 professional job roles across 24 categories — from software engineering and data science to legal, finance, marketing, and supply chain. Each role has structured prompt templates with context fields, task types, and output formats. It's the prompt library your AI app shouldn't have to build from scratch.
Base URL: https://godle.app/api/v3 — No API key. No signup. CORS open. CDN-cached. 185 roles, 1,741 templates, 12 workflows, MCP manifest, A2A agent card. JavaScript SDK available at /api/v3/godle-sdk.js.
API Endpoints
All endpoints return JSON. All are GET requests. All are served from CDN with aggressive caching.
| Endpoint | Description |
|---|---|
| /api/v3/index.json | API manifest — all endpoints, stats, protocol adapters |
| /api/v3/roles.json | Index of all 185 roles with slugs, categories, template counts |
| /api/v3/roles/{slug}.json | Full role definition with all templates and context fields |
| /api/v3/categories.json | 24 job categories with role listings |
| /api/v3/capabilities.json | All 1,741 templates in a flat index (good for search) |
| /api/v3/workflows.json | 12 multi-step AI workflow definitions |
| /api/v3/team-packs.json | 8 curated role bundles by team type |
| /api/v3/godle-sdk.js | UMD JavaScript SDK (94KB, zero dependencies) |
| /.well-known/mcp.json | MCP tool manifest for AI agent integration |
| /.well-known/agent.json | A2A agent card (Google Agent-to-Agent protocol) |
Quickstart: Fetch a Role
Get all templates for a software engineer in one fetch:
Role Slugs: All 185
Fetch the roles index to get all available slugs programmatically:
A few examples of what's available: software-engineering, data-science, machine-learning-engineering, ux-design, product-management-core, financial-planning-analysis, regulatory-compliance, enterprise-field-sales, talent-acquisition, customer-success-management, security-operations, sre-devops.
Template Structure
Each template in a role has a consistent schema:
The contextFields array tells you exactly what variables to inject for a complete prompt. The taskType is useful for filtering: values include code-review, writing, analysis, planning, communication, research, and more.
Using the JavaScript SDK
For more complex integrations, the SDK handles fetching, caching, variable interpolation, and adapter setup for Anthropic and OpenAI. The full source, docs, and examples are available on GitHub (MIT license).
MCP Integration
Godle exposes an MCP (Model Context Protocol) manifest at /.well-known/mcp.json. This means any MCP-compatible agent or tool (Claude Desktop, custom agents, etc.) can discover and use Godle's prompt templates as tools automatically — without any custom integration code.
What Developers Are Building With It
The static API is intentionally unopinionated. Common uses we've seen:
- Internal tools: Prompt pickers embedded in Slack bots or internal dashboards, surfacing the right template based on what the user is working on
- AI agent systems: Using the role templates as the system prompt library for multi-role agent workflows — each agent gets its role's context and templates
- Chatbot persona seeding: Using role definitions to configure domain-specific chatbots with appropriate expertise framing
- Prompt testing frameworks: Using the template library as a benchmark dataset for evaluating LLM prompt sensitivity and output quality across roles
- No-code AI tools: Embedding the role selector in no-code AI builder platforms as the prompt foundation layer
Because the entire API is static JSON, you can download and bundle it with your application for fully offline use. The roles.json + individual role files total ~15MB uncompressed. The capabilities.json flat index is the most useful single file for search-based use cases.
Rate Limits & SLA
Layer 1 (the static API described in this post) has no rate limits — it's a CDN. Vercel's CDN serves it from edge nodes globally with effectively unlimited throughput for read requests. There is no authentication and no usage tracking on this layer.
If you need stateful sessions, task execution, or SSE streaming (Layer 2), that runs on Cloudflare Workers at orchestration.godle.app and requires a Bearer token.
Try the prompt generator
The API powers Godle's interactive prompt generator — 185 roles, no signup, free.
⚡ Try Godle FreeOr fetch the API directly: godle.app/api/v3/index.json