Scenarios

A scenario has several AIs work one after another on the same request. Each AI has a role: for example, one AI splits the work, others do it, a last one reviews it. You draw this circuit once, then reuse it.

Examples: have a pull request reviewed by a developer, a tech lead and a security lead, or study a decision from several angles.

How it is built

A scenario is a diagram made of nodes joined by arrows.

• A node is a step given to an AI. It has a role, instructions, a tool (Claude Code or Codex) and a model. An ask node asks you a question and waits for your answer before going on.
• An arrow links a node to the one that works after it. The node above is the "parent". Each node receives its parent's result.
• The leader is the first node: it receives your request.
• Groups arrange nodes on screen. A node must belong to a group, but groups do not change the order of work.

The two modes

Mode	How it works
Hierarchical (default)	The leader splits the request and gives an assignment to each node under it. It can review their work and hand it out again. Each node has a single parent.
Topological	Nodes work from top to bottom, without the leader splitting the work. A node can have several parents: it then waits until they have all finished, and receives each one's result.

What a node receives

A node always receives your request and its parent's result. In topological mode, if it has several parents, it receives each one's result.

The Additional ancestor outputs to inject setting also gives it the result of the chosen nodes, even if they are not its parents. Example: a review node at the bottom of the diagram also receives the analysis made at the start. This setting works in both modes.

Creating a scenario

1. Open Settings → Scenario (in the "AI" group), then click New Scenario.
2. Start from a preset or from an empty diagram. You can also click Generate with AI and describe the scenario in a few sentences: Runeya draws the diagram, which you then adjust.
3. Add a group, then your nodes with their role and instructions. Link them with arrows.

A scenario has at most 500 nodes and 100 groups. A node's instructions are at most 50,000 characters.

Presets

A preset is a starting scenario, ready to adapt. Picking one in the editor creates the groups, the nodes, their links and their instructions, then lays out the graph automatically. Names and instructions are in the interface language at the time you apply the preset, and roles keep that language when the scenario runs.

Preset	Mode	What it does
Empty company	hierarchical	A director alone, with ten positions available: Manager, Engineer, Designer, Analyst, QA, DevOps, Product, Writer, Researcher and Other.
Full company	hierarchical	The director asks the planner for a written plan, then hands the tasks to the Front, Back, Security, QA, Documentation and DevOps services. Each service lead splits the work among its members, reviews it and sends it back when it does not meet the specification.
Feature delivery	topological	Product framing, design, build, then testing and documentation in parallel.
Bug investigation	topological	Triage, a question about reproduction steps, root cause search, fix with a test, verification.
Research report	topological	Research questions, facts and options gathered in parallel, analysis, report.
Code review	hierarchical	A lead reviewer splits the review between security, correctness and maintainability, then merges the findings.

In hierarchical presets, the director, the planner, managers and service leads do not write code: they split, delegate and review.

Running a scenario

1. In the chat, click New conversation.
2. In the "Workflows" section, click the scenario.
3. In the start window, write your request. You can also pick the runner, the agent and the model.
4. Click Start.

A conversation is created and the scenario starts. Each step shows in it as it happens. When an ask node waits for your answer, Runeya sends you a notification, and the work resumes as soon as you answer.

The Stop button ends the scenario. The conversation keeps what was done, but a stopped scenario cannot resume.

Runeya saves the log and the final summary of each run in the scenario-runs/ folder, with your other data.

What a node can do

A node runs Claude Code or Codex in the folder of the conversation's project. It has the same tools as a chat conversation:

• its CLI's tools: read and edit files, run commands;
• Runeya's tools: see the state of services, read their logs, start or stop them, read and comment on kanban cards;
• the active environment's variables marked "AI inject".

A node can therefore act on your services. State in its instructions what it is allowed to do.