Describe the work. Agents build the agents with you.
When nothing in the catalog fits, describe the process. Celara's own agent graphs interview your people, map how the work really happens, and build a governed agent graph with you. What they build runs on the same rails as every installed solution.
- 01Discovery interview
Your people, topic by topic
transcripts - 02Org analysis
Who does what, with which tools
org graph - 03PRD
Requirements from what people said
reviewed diff - 04Evals
How success is checked, agreed first
eval suite - 05Project
Created in Studio from the PRD and evals
agent graph - 06Live on Fabric
Runs under the evals, as gates
active agents
It asks what a good analyst would ask. Then it asks someone else.
An agent graph interviews the people who do the work, one topic at a time. It looks for the rule behind each habit, and for the places where two people describe the same step differently.
- Topic by topic. Each stakeholder covers the topics they own. Interviews can be paused and resumed.
- Disagreements surface early. When answers conflict, the interview flags it, so the conflict is settled before it becomes a rule.
- Answers become drafts. What people say feeds the requirements and the first draft of the articles that will govern the work.
Who works with whom, and which agents they use.
Org analysis maps people, roles, teams and agents, and how they connect across business units. It shows who a new agent graph will touch before anyone builds it.
An org map of four business units, Sales, Legal & Compliance, Finance and Operations, with their people, their agents and the bridges between units. A proposed client-onboarding graph is added, and the map shows it touches three units: Maya in Sales as the account owner, Priya in Compliance for review, Inês in Finance for billing, and the Contract review agent. Operations is not touched.
Touches 3 units, 3 people and 1 agent.
- Sales: Maya, as the account owner
- Legal & Compliance: Priya reviews doubtful clients, and the Contract review agent checks the engagement letter
- Finance: Inês sets up billing
Operations isn’t touched. Everyone it affects is known before anyone builds it.
From what people said to a project you can build.
Each step drafts a document your team reviews and approves. The evals are agreed before anything is built, and they become the gates in Fabric.
Purchase approvals
Agreed before the build
What will be active
Sample documents. Billing follows what's active: agents, graphs, tools and skills. Token use is never charged.
The plan becomes a graph while you watch.
Studio is where the project gets built. Tell it what the graph should do. It asks a few questions, shows you the plan, and builds the agent graph on the canvas, with the eval gates and approval steps already in place.
A replay of Studio. Someone asks for a check on every new client. Studio asks which risk level needs a person, shows a four-step plan and builds the graph: sanctions, adverse-media and ownership screens, a risk score, and a compliance officer for doubtful cases. A test run with a sample client stops at the compliance officer.
Start from one sentence.
Studio opens on an empty canvas and a chat. Say what the graph should do, or add the input yourself. When the graph is ready, run it in the playground and publish it to Fabric.
How Fabric runs it →Build it yourself, or with our engineers.
Most teams build in Studio on their own. If you need them, our forward deployed engineers work alongside your team to tailor a solution to your systems, or to build one the catalog doesn't have yet.
Your first session is the interview.
Bring the people who know how one process really works. Nothing needs to be prepared.