# agents.txt version: 2.1
# AINative Studio — Agent Discovery File
# https://www.ainative.studio/agents.txt
# Standard: agents.txt (analogous to llms.txt / robots.txt for AI agents)

## Platform

Name: AINative Studio
Description: AI-native development platform with serverless vector database, multi-agent orchestration, MCP server hosting, and agent-to-user interface protocol.
URL: https://www.ainative.studio
API Base: https://api.ainative.studio
OpenAPI Spec: https://api.ainative.studio/docs
OpenAPI JSON: https://www.ainative.studio/openapi.json
Contact: toby@ainative.studio
Founded: 2025-09
Location: Santa Cruz, California

## Agent Compatibility

This platform is built for agents. All APIs accept programmatic access via API key.
Zero-friction onboarding: curl https://api.ainative.studio/v1/chat/completions -H "X-API-Key: $KEY" -d '{"model":"default","messages":[{"role":"user","content":"hello"}]}'

## Products

### ZeroDB
Type: Serverless vector database
Description: Store embeddings, run semantic search, manage NoSQL tables, event streams, file storage, a knowledge graph, and RLHF/preference data — one API for your full agent data layer. No infrastructure to manage.
Docs: https://docs.ainative.studio/zerodb
NPM: npm install zerodb-mcp
PyPI: pip install zerodb-mcp
MCP Server: ainative-zerodb-memory-mcp (6 tools)
Primitives: memory (remember/recall), vector search, embeddings, NoSQL tables, event streams, file storage, knowledge graph (entities/centrality/contradictions), RLHF & preference data
RLHF API: POST /api/v1/public/zerodb/{project_id}/database/rlhf/interactions (log), .../feedback (score), .../interactions/compare (pairs), .../export (training-ready dataset)

### Lakehouse
Type: Serverless columnar analytics (free queries)
Description: Run unlimited SQL over your Parquet with in-memory DuckDB. Queries are free and unmetered — no cluster, no compute bill. Pay only for storage ($0.15/GB, 10 GB free). Part of the ZeroDB platform: one agent write becomes a row, a bulk-queryable record, a semantic memory, and a knowledge-graph node — no ETL.
Landing: https://ainative.studio/products/lakehouse
Docs: https://docs.ainative.studio/docs/api/zerodb-lakehouse
Query: POST /api/v1/projects/{project_id}/database/lakehouse/query  { "sql": "SELECT ..." }  (X-API-Key header)
Ingest: POST /api/v1/projects/{project_id}/database/lakehouse/ingest  { "sensor_type": "string", "records": [...] }
Pricing: queries free; storage Free 10GB / Pro $0.15/GB / Enterprise $0.10/GB

### Agent Swarm
Type: Multi-agent orchestration
Description: 6 specialized autonomous agents — CTO/Architect, Frontend, Backend, DevOps, QA, and Testing. 46+ tools. Handles PRD to production.
Docs: https://docs.ainative.studio/agent-swarm

#### Cody Jackson (CTO / Architect)
Role: Lead architect and technical decision-maker
Capabilities: System design, code review, API design, database schema, architecture decisions
Tools: code_review, architecture_plan, api_design, schema_design, dependency_audit, security_review
Input: { "task": "string", "context": { "repo": "string", "constraints": [] } }
Output: { "plan": "string", "files_to_modify": [], "estimated_complexity": "low|medium|high" }

#### Forrest Kinkade (SRE / DevOps)
Role: Infrastructure, deployment, and reliability engineering
Capabilities: CI/CD pipeline config, Docker/Railway deployment, monitoring setup, incident response
Tools: deploy_railway, configure_ci, setup_monitoring, check_health, manage_secrets, scale_service
Input: { "task": "string", "environment": "staging|production", "service": "string" }
Output: { "status": "string", "deployment_url": "string", "logs": "string" }

#### Frontend Agent
Role: UI/UX implementation and component development
Capabilities: React/Next.js components, Tailwind styling, accessibility, responsive design
Tools: create_component, style_component, add_animation, accessibility_audit, responsive_check
Input: { "task": "string", "framework": "react|nextjs|vue|svelte", "design_spec": {} }
Output: { "files_created": [], "component_tree": {}, "accessibility_score": "number" }

#### Backend Agent
Role: API development and business logic
Capabilities: FastAPI endpoints, database queries, authentication, service integration
Tools: create_endpoint, write_query, add_middleware, integrate_service, generate_schema
Input: { "task": "string", "stack": "fastapi|express", "database": "postgres|zerodb" }
Output: { "endpoints_created": [], "migrations": [], "test_coverage": "number" }

#### QA Agent
Role: Quality assurance and test strategy
Capabilities: Test plan creation, edge case identification, regression testing, performance testing
Tools: create_test_plan, identify_edge_cases, run_regression, performance_benchmark, coverage_report
Input: { "task": "string", "test_type": "unit|integration|e2e|performance" }
Output: { "test_plan": {}, "test_files": [], "coverage": "number", "issues_found": [] }

#### Testing Agent
Role: Automated test execution and CI validation
Capabilities: Run test suites, validate builds, check linting, verify deployments
Tools: run_tests, validate_build, check_lint, verify_deployment, generate_report
Input: { "task": "string", "suite": "string", "environment": "string" }
Output: { "passed": "number", "failed": "number", "report_url": "string" }

### MCP Server Hosting
Type: Model Context Protocol infrastructure
Description: Deploy and manage 14+ pre-built MCP servers with auto-scaling. One-click deployment, no Docker required.
Available servers: GitHub, PostgreSQL, Memory, Slack, Gmail, Filesystem, Brave Search, Puppeteer, SQLite, Fetch, EverArt, Sentry, AWS Knowledge Base, Shadcn UI
Manifest: https://www.ainative.studio/.well-known/mcp-servers.json
Docs: https://docs.ainative.studio/mcp-hosting

### A2UI Protocol
Type: Agent-to-User Interface standard
Description: Protocol enabling agents to generate dynamic, interactive UI components at runtime. No frontend code required.
NPM: @ainative/ai-kit-a2ui-core
Install: npm install @ainative/ai-kit-a2ui-core
Docs: https://docs.ainative.studio/a2ui

## AI Models Catalog

Serverless AI models (text/coding/audio/video/image/embeddings) via a single OpenAI-compatible endpoint.
The catalog is dynamic — query the live registry for the authoritative, current list:
Catalog endpoint: GET https://api.ainative.studio/api/v1/public/ai-registry/models (no auth)
Chat endpoint: POST https://api.ainative.studio/v1/chat/completions

# NOTE: api_id values below are from the live registry (GET /api/v1/public/ai-registry/models).
# The catalog changes over time — always treat that endpoint as the source of truth.

### Text / Coding Models
name: Qwen3 32B | api_id: qwen3-32b | tool_calling: true
name: Qwen3 14B | api_id: qwen3-14b | tool_calling: true
name: Qwen3 8B | api_id: qwen3-8b | tool_calling: true
name: Claude 3.5 Sonnet | api_id: claude-3-sonnet | provider: anthropic
name: GPT-4 | api_id: gpt-4 | provider: openai
name: Google Gemini 2.0 Flash | api_id: gemini-flash | provider: google
name: Cohere Command A | api_id: cohere-command | provider: cohere
name: Mistral Medium | api_id: mistral-medium | provider: mistral
name: Llama 4 Maverick 17B | api_id: llama-4-maverick
name: NousCoder | api_id: nous-coder | specialty: coding

### Audio Models
name: Kokoro-82M | api_id: kokoro-82m | type: tts
name: MeloTTS | api_id: melotts | type: tts
name: MiniMax TTS Sync | api_id: minimax-tts-sync | type: tts
name: Text-to-Speech (OpenAI) | api_id: openai-tts | type: tts
name: MiniMax Music 2.5 | api_id: minimax-music-v2 | type: music
name: Whisper Transcription | api_id: whisper-transcription | type: asr
name: Whisper Translation | api_id: whisper-translation | type: asr

### Video Models
name: Alibaba Wan 2.2 I2V 720p | api_id: alibaba-wan-22-i2v-720p | type: image-to-video
name: CogVideoX-2B | api_id: cogvideox-2b | type: text-to-video
name: MiniMax Hailuo 2.3 | api_id: minimax-hailuo | type: video
name: MiniMax Hailuo 2.3 Fast | api_id: minimax-hailuo-fast | type: video
name: Seedance I2V | api_id: seedance-i2v | type: image-to-video
name: Sora2 | api_id: sora2-i2v | type: image-to-video
name: Text-to-Video | api_id: text-to-video-model | type: text-to-video

### Image Models
name: MiniMax Image-01 | api_id: minimax-image-01 | type: text-to-image
name: Qwen Image Edit | api_id: qwen-image-edit | type: image-edit

### Embedding Models
name: BGE Small EN v1.5 | api_id: bge-small-en-v1-5 | dims: 384
name: BGE Base EN v1.5 | api_id: bge-base-en-v1-5 | dims: 768
name: BGE Large EN v1.5 | api_id: bge-large-en-v1-5 | dims: 1024

### Tool Calling Notes
- All models: POST /v1/chat/completions {"model":"<api_id>","messages":[...],"tools":[...]}
- Native tool calling: qwen3-32b, qwen3-14b, qwen3-8b
- Requesting tools on a non-tool model routes to a tool-capable Qwen3 automatically
- Always confirm a model's current api_id and capabilities via GET /api/v1/public/ai-registry/models

## Zero-Human Provisioning (No Signup Required)

Agents can provision a working ZeroDB project with API key in one request.
No signup, no credit card, no human required.

### Instant Database
POST https://api.ainative.studio/api/v1/public/instant-db
Description: Provision a ZeroDB project instantly. Returns API key, project ID, connection string, and MCP config.
Auth: None required.
Rate limit: 5 per IP per day.
Response: { "api_key": "tmp_...", "project_id": "uuid", "base_url": "https://api.ainative.studio", "mcp_url": "npx zerodb-cli connect <project_id>", "connection_string": "postgresql://...", "expires_at": "72 hours from now" }
Limits: 10,000 vectors, 1,000 memory records, 50MB storage, 50 chat credits.
Claim URL: Returned in response — human can claim to convert to permanent account.

Example (agree_terms:true is REQUIRED — returns HTTP 451 without it):
  curl -X POST https://api.ainative.studio/api/v1/public/instant-db -H "Content-Type: application/json" -d '{"agree_terms": true}'

## API Endpoints (Agent-Relevant)

Authentication: X-API-Key header required for all endpoints.

### Chat Completions
POST https://api.ainative.studio/v1/chat/completions
Description: OpenAI-compatible chat completions endpoint.
Body: { "model": "string", "messages": [...], "stream": false }

### Vector Upsert
POST https://api.ainative.studio/api/v1/projects/{project_id}/database/vectors/upsert
Description: Upsert embeddings into ZeroDB. Auto-embeds text if no vector provided.
Body: { "namespace": "string", "id": "string", "vector_embedding": [...], "metadata": {} }

### Vector Search
POST https://api.ainative.studio/api/v1/projects/{project_id}/database/vectors/search
Description: Semantic similarity search across a ZeroDB namespace.
Body: { "namespace": "string", "vector_embedding": [...], "top_k": 10 }

### Project Create
POST https://api.ainative.studio/v1/projects
Description: Create a new AINative project workspace.
Body: { "name": "string", "description": "string" }

### Agent Task (Agent Swarm)
POST https://api.ainative.studio/api/v1/public/agent-swarm/tasks
Description: Submit a task to the Agent Swarm for autonomous execution.
Body: { "task": "string", "context": {} }

### Memory Store
POST https://api.ainative.studio/v1/public/memory/v2/remember
Description: Store a memory for an AI agent session.

### Memory Recall
POST https://api.ainative.studio/v1/public/memory/v2/recall
Description: Retrieve relevant memories for an agent session.

### Knowledge Graph — GraphRAG
POST https://api.ainative.studio/api/v1/public/memory/v2/graph/graphrag
Description: Hybrid vector + knowledge-graph search over your memory.

### Knowledge Graph — SPARQL (W3C RDF)
POST https://api.ainative.studio/api/v1/public/memory/v2/graph/sparql
Description: Read-only SPARQL 1.1 (SELECT/ASK/CONSTRUCT) over your namespace's RDF knowledge graph. Row-capped; writes rejected.
Body: { "query": "SELECT ?s ?p ?o WHERE {?s ?p ?o} LIMIT 25" }

### Knowledge Graph — Resolve / Dereference
GET https://api.ainative.studio/api/v1/public/memory/v2/graph/vault-ld/resolve?format=turtle
Description: Content-negotiated dereference of your namespace graph (turtle | jsonld | markdown). Entity IRIs resolve at https://memory.ainative.studio/{tenant}/{namespace}/{kind}/{local}.

### Knowledge Graph — Export / Import (Vault-LD)
GET  https://api.ainative.studio/api/v1/public/memory/v2/graph/export/vault-ld?format=vault-ld  (or format=turtle)
POST https://api.ainative.studio/api/v1/public/memory/v2/graph/import/vault-ld
Description: Lossless RDF export/import. Bi-temporal edges (OWL-Time) + PROV-O provenance. DPROD descriptor at /graph/vault-ld/dprod.

## SDKs

TypeScript/JS: npm install @ainative/sdk
Python:        pip install ainative
Go:            go get github.com/AINative-Studio/zerodb-go-sdk
MCP (Node):    npx ainative-zerodb-memory-mcp
MCP (Python):  pip install zerodb-mcp

## Developer Program

Developers can build apps on the AINative platform and earn revenue from their customers' API usage.
Markup range: 0–40% on API calls
Platform fee: 5% of developer earnings
Payouts: Weekly via Stripe Connect (minimum $10)
Sign up: https://www.ainative.studio/developers

## Getting Started

- Agent Quickstart Guide: https://www.ainative.studio/getting-started/agents
- General Getting Started: https://www.ainative.studio/getting-started
- API Documentation: https://docs.ainative.studio
- Changelog: https://www.ainative.studio/changelog

## Rate Limits

- Free tier: 100 requests/minute, 1,000/hour, 10,000/day
- Pro tier: 500 requests/minute, 5,000/hour, 50,000/day
- Business/Enterprise: Custom limits
- Rate limit headers: X-RateLimit-Limit, X-RateLimit-Remaining, X-RateLimit-Reset

## Discoverability Files

llms.txt:        https://www.ainative.studio/llms.txt
agents.txt:      https://www.ainative.studio/agents.txt
sdks.txt:        https://www.ainative.studio/sdks.txt
ai-plugin.json:  https://www.ainative.studio/.well-known/ai-plugin.json
mcp-servers:     https://www.ainative.studio/.well-known/mcp-servers.json
openapi.json:    https://www.ainative.studio/openapi.json
sitemap:         https://www.ainative.studio/sitemap.xml

## Privacy Terms (IEEE 7012-2025)

AINative Studio publishes machine-readable privacy terms per IEEE 7012-2025 (My Terms standard).
Agents MAY verify these terms before provisioning an account or submitting payment.

MyTerms: https://ainative.studio/.well-known/myterms.json
Agreement type: SD-BASE (ongoing service relationship, no data sale)
Sells data: false
Third-party sharing: analytics, error-tracking only
Retention: 365 days
Machine-readable: true
Human policy: https://ainative.studio/privacy