VIABLE Lab Developer Platform

One API for chat, VETTING, and multimodal AI, built for learning.

A single model-first API with OpenAI-compatible request shapes. Ship tutoring assistants, classroom tools, and learning-science experiments with verified responses, agent workflows, scoped keys, and request-level tracing, end to end.

Get an API key Read the quickstart

OpenAI-compatible Server-side keys Request-level tracing

curl https://api.viablelab.org/v1/chat/completions \
  -H "Authorization: Bearer $VIABLE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "viable-2",
    "messages": [
      { "role": "user", "content": "Explain photosynthesis to a 5th grader." }
    ]
  }'

Every call is traceableRequest ID, model, token usage, and estimated cost land in your dashboard.

BuildIntelligent tutoringAutomated feedbackAnswer verificationAgent workflowsLearning analyticsResearch experiments

Capabilities

Everything you need, behind one base URL.

Chat, verified educational responses, agent orchestration, multimodal input, embeddings, media generation, and audio share a single catalog, one auth model, and consistent logging.

Chat & reasoning

OpenAI-compatible chat completions with model-first routing, streaming, and multimodal content arrays.

/v1/chat/completionsStreaming

VETTING

A verification layer over any chat model. Draft, verify against guidance or an answer key, and retry until it passes.

/v1/chat/vetting/v1/vetting/full

Agent orchestration

Build visual or JSON workflows that branch, loop, call models, generate assets, and publish as one stable endpoint.

/v1/agents/{slug}/runsCanvas + code

Multimodal input

Send text, images, files, and PDFs as content parts to models that support them, with provider conversion handled for you.

ImagesPDFFiles

Embeddings

OpenAI-compatible vectors for retrieval, semantic search, clustering, and reranking pipelines.

/v1/embeddings

Image generation

Text-to-image generation with familiar Images-style request shapes across supported providers.

/v1/images/generations

Speech & transcription

Generate speech from text or transcribe uploaded audio with OpenAI-compatible audio endpoints.

/v1/audio/speech/v1/audio/transcriptions

Signature capability

VETTING: generation you can actually trust.

VETTING isn't a model. It's a verification architecture layered on top of any chat model. A model drafts a response, a verification pass checks it against your guidance or answer key, and it retries until it passes or hits your attempt limit.

Guide without giving it awayTutor-style answers that won't reveal the homework solution.

Answer-key verificationCheck responses against reference answers with structured, auditable attempt records.

Mix providersUse different chat and verification models across OpenAI, Google, or VIABLE Lab.

Explore VETTING

1DraftChat model generates a candidate response.

2VerifyA verification pass scores it against guidance or answer key.

3Pass or retryReturn when it passes, or retry with feedback up to your limit.

Built for humans and agents

An agent-native developer workflow.

Whether you're wiring up a classroom tool, a product feature, or a research pipeline, you can work directly from the API reference or use the maintained agent guide to help an AI assistant draft endpoint calls and agent workflow JSON.

Create a scoped key

Generate a server-side key in the dashboard with rate, token, cost, and model limits before it reaches production.

Pick the right shape

Choose direct endpoints for simple calls, or publish an agent workflow when the product needs branching or repeated model steps.

Hand off to your AI agent

Copy the maintained guide into Claude Code, Codex, Cursor, or Gemini CLI so it can work from real endpoint and agent-graph context.

Trace every request

Review status, model, usage, estimated cost, request IDs, and logged text payloads for research replay.

Agent context

Building with a coding agent?

Drop the maintained API guide into Claude Code, Codex, Cursor, Gemini CLI, or another AI-assisted workflow. Your agent can use direct endpoints when they fit, or draft VIABLE Lab agent workflow JSON when you need orchestration.

Codex Claude Code Gemini CLI Cursor Windsurf

Open the agent guide

I am building with the VIABLE Lab API. Read https://developers.viablelab.org/agents/viablelab-api.md as the source of truth. Prefer direct endpoints for one-step chat, VETTING, embeddings, image, speech, transcription, or encryption; suggest Dashboard Agents only when a product flow needs branching, loops, or multiple coordinated model/tool calls behind one endpoint. Auth uses Bearer VIABLE Lab API keys with endpoint scopes such as model:request, utils:crypto, and agents:execute. For published agent runs, call POST /v1/agents/{slug}/runs and persist conversation_id per end-user thread + agent slug; run_id is one execution and request_id is for log/support correlation. Treat any direct-endpoint or provider-level session state as separate from this agent conversation_id.

Developer operations