Auglet for AI agents
Auglet is the install layer for AI agents — a marketplace for Prompt Sets, Skills, Tools, Harnesses, and Other Agents.
A2A Agent Card
- Agent Card: https://auglet.com/.well-known/agent-card.json
- Alias: https://auglet.com/.well-known/agent.json
- Mode: operable — public discovery skills callable via A2A POST
- Extended Agent Card (auth required): https://fjozfegsxseblmpplcrb.supabase.co/functions/v1/a2a/agent-card
- Human page: https://auglet.com/agents
A2A endpoint
POST https://fjozfegsxseblmpplcrb.supabase.co/functions/v1/a2a
Content-Type: application/json
{
"skillId": "ask_about_auglet",
"message": {
"parts": [{ "type": "text", "text": "What is Auglet?" }]
}
}
Poll task status:
GET https://fjozfegsxseblmpplcrb.supabase.co/functions/v1/a2a/tasks/{taskId}
Public skills (no API key)
| Skill ID | Name | Description |
|---|---|---|
get_business_profile |
Business Identity Profile | Returns Auglet's published business identity: legal entity, service area, produc… |
ask_about_auglet |
Ask About Auglet | Deterministic FAQ Q&A about Auglet using published corpus only. No LLM. Required… |
ask_auglet_agent |
Ask Auglet Agent | Grounded AI-assisted natural-language Q&A about Auglet. Answers strictly from pu… |
browse_catalog_index |
Public Catalog Index | Returns the public marketplace catalog index (slug, title, category, price, URL)… |
get_contact_options |
Contact and Support Routing | Returns published contact channels, use cases, and routing guidance. Does not ex… |
Each skill also has backing pointers in the Agent Card for direct MCP/HTTP/static invocation.
Phase 1 A2A workflows (operable)
These four skills execute over POST https://fjozfegsxseblmpplcrb.supabase.co/functions/v1/a2a. Tasks are persisted when the task store is configured; poll GET /tasks/{taskId} for artifacts and history.
| Skill | Required input | Artifacts | Missing input |
|---|---|---|---|
get_business_profile |
Any message part | business_profile (JSON) |
input-required |
ask_about_auglet |
Text part with question | faq_answer (JSON), faq_answer_text |
input-required |
browse_catalog_index |
Optional JSON { query?, category?, limit? } |
catalog_index (JSON) |
input-required on invalid filters |
get_contact_options |
Any message part | contact_options (JSON) |
input-required |
Example — catalog filter:
{
"skillId": "browse_catalog_index",
"message": {
"parts": [{
"type": "json",
"content": { "category": "skills", "limit": 10 }
}]
}
}
Authenticated skills on the extended Agent Card return auth-required without an API key. start_checkout returns input-required until human approval; checkout execution remains deferred.
Other integration surfaces
| Surface | URL |
|---|---|
| MCP server | https://fjozfegsxseblmpplcrb.supabase.co/functions/v1/mcp-server |
| REST API | https://fjozfegsxseblmpplcrb.supabase.co/functions/v1/api-v1 |
| MCP setup guide | https://auglet.com/mcp |
| Agent readme | https://auglet.com/agent-readme.md |
| Public catalog | https://auglet.com/catalog.json |
| External agent registry | https://auglet.com/agents/registry |
| Registry catalog JSON | https://fjozfegsxseblmpplcrb.supabase.co/functions/v1/a2a-registry/catalog/agents.json |
Authentication
Public skills require no API key. Authenticated skills require an API key from Creator Studio → API Console with the appropriate scopes. Fetch your scope-filtered skill list from the extended Agent Card:
GET https://fjozfegsxseblmpplcrb.supabase.co/functions/v1/a2a/agent-card
X-Api-Key: mk_live_...
Invalid or missing credentials return 401. Valid credentials without required scope return 403 with task status rejected.
Human approval
Agents must obtain human confirmation before: start_checkout, download_purchase, submit_support_feedback. Server returns task status input-required with a one-time approvalToken; resubmit with taskId and metadata.approvalToken.
Do not invent
Do not invent phone numbers, street addresses, business hours, customer reviews, or certifications for Auglet.