Create with the agent
Describe a decision in plain words and let the Dcision agent build it in the visual builder — then keep refining it by chat while you watch every change.
Goal: go from an idea to a working decision without writing the schema yourself. The agent writes the questions, rules and destinations; you see each change land in the Visual builder and can still edit anything by hand.
Start a decision by chat
- Decisions → New decision. The first block is Describe the decision.
- Write what you want to decide — the input, the answer you need and what should happen next. For example: "Read each support message, pick the team that should answer (billing, technical or account) and how urgent it is; send urgent cases and angry customers to a person." The example chips fill in common cases.
- Create with the agent. Dcision creates the decision and opens the Visual builder on the Chat tab.
You watch the agent work: its reply streams in as it writes, a status line shows what it's doing ("Reading the draft", "Saving the draft"), each saved change becomes an activity line ("Added the urgency question · rev. 3"), and the canvas, the inspector and the Editor tab follow every revision.
When the agent needs a choice only you can make — which market, where results should go — it asks a question with 2–4 options (one question at a time). Click one (or tick several and Send) and it carries on; you can also write another answer. Answered questions keep the chosen option marked for the whole team. The agent never asks for API keys or passwords, and the chat refuses messages that contain a Dcision API key.
Prefer a fixed starting point? Below the chat box, Or start from a template keeps the templates and the blank decision.
Keep refining
In the builder, the left column has two tabs: Chat and Components (the palette). Ask for changes in plain words:
- "Add a question about urgency."
- "When urgency is urgent, escalate and answer with a fixed message."
- "Send hot leads to our n8n workflow."
Each change is saved as a new revision of the draft. The agent never deploys: test in the Playground and click Deploy when you're happy. New conversation starts a fresh chat (the decision stays as it is).
Editing by hand at the same time
Your own edits are saved first when you send a message, so the agent starts from your latest draft (if your draft has problems, fix them or discard your edits before sending). If you edit while the agent is working and it saves a new revision, the chat shows The agent changed the draft: Use the agent's version discards your unsaved edits; Keep mine keeps them, and saving replaces the agent's changes.
What the agent can and can't do
| It can | It can't |
|---|---|
| Read and rewrite the draft of this decision: state, questions, composites, rules, destinations, settings, name, description and slug (until the first deploy) | Deploy, delete, disable or touch other decisions |
| Validate the schema and run the saved draft once with an example (like the Playground) | Read or create secrets and API keys — destinations reference them as {{secrets.NAME}}, set by an admin |
| Start from a template | Remember anything from one conversation for another |
Limits
- Messages up to 4,000 characters; one agent reply at a time per decision.
- Up to 30 messages per hour and 200 per day per workspace, and 20 per hour per person.
- Viewers can read the conversation; members, admins and owners can chat.
MCP server
The Dcision MCP server at https://api.dcision.io/mcp — authentication, the fourteen tools with their inputs and outputs, errors, and setup for Claude Code, Cursor, VS Code, Claude Desktop, Codex, Gemini CLI and curl.
Lead qualification
Score purchase intent, set a priority, rank with a lead score and route inbound leads to sales, SDRs or nurture — and block spam — with the Lead Qualification template.