AI as a teammate, not a chatbot.
Three scoped agents do real, structured work: Deep Research proposes outlines and finds sources; Curator selects the best; Generation drafts per-channel variants.
Why this matters
Chat-style AI is a dead end for serious content work. You paste in a question, get a paragraph, and have no idea where the information came from or what to do with it. Real AI work is scoped: it has a job, an input, a method, and a structured output that plugs into the next step of your pipeline.
Four steps, no friction.
- 1
Pick an agent
Deep Research, Curator, or Generation. Each has a defined role and a structured output. No free-form chat.
- 2
Scope the run
Pass the agent a Topic, a Source Library, or a research question. The agent only acts on the scope you give it — nothing else.
- 3
Watch the run
Every step is logged: sub-questions generated, sources evaluated, sources selected. You see what the agent did and why.
- 4
Use the output
Research brief becomes a draft outline. Curator’s selections become the Source Library. Generation becomes per-channel variants ready to publish.
What you get out of the box.
- Three agents: Deep Research, Curator, Generation
- OpenAI-compatible, Gemini, Anthropic, and Groq providers
- Per-Project agent config — choose your model, temperature, and limits
- Run logs with full input and output JSON, kept for 90 days
- Per-Project token quota with visible usage and overage warnings
- Bring your own API key — your tokens are billed to your account
Screenshot of the Contents Lab AI agent run view showing the Deep Research agent progress and source selections.
Agents are scoped, not open-ended.
The problem. A chatbot answer is just a paragraph. You cannot build on it.
The Contents Lab way. Each agent has a fixed output schema. Deep Research returns an outline + a ranked source list. Curator returns a selection. Generation returns a draft + per-channel variants. Every output plugs into the next step.
- Outputs are JSON, not free text
- Each agent run is idempotent — re-run with the same input, get the same shape
- Agent configs are versioned; roll back to a prior config in one click
Bring your own provider and key.
The problem. Lock-in to one AI vendor is expensive and slow to change.
The Contents Lab way. Configure any OpenAI-compatible endpoint (OpenAI, Together, Groq, Ollama) plus Gemini and Anthropic. Swap providers per agent or per run.
- Per-provider API keys stored encrypted in the database
- Per-Project default provider, overridable per agent
- Cost tracking by request so you see exactly what each run cost
Try AI as a teammate, not a chatbot today.
Free plan, no card required. Cancel anytime.