.bot/ directory inside your project repository.
Launching the dashboard
Start the runtime and the dashboard together from your project root:dotbot go brings up the per-project HTTP runtime and then the dashboard server. The dashboard auto-selects an available localhost port, writes it to .bot/.control/ui-port, and prints the URL it bound to. Open that URL in any browser to access the dashboard.
To pin the dashboard to a specific port, pass --port. To open the browser automatically, pass --open:
dotbot go attaches to it rather than starting a second one — the model is one dashboard per project, showing all runs.
dotbot go starts both the runtime and the dashboard. If you only need the HTTP runtime — for diagnostics or tests — use dotbot serve instead, which runs the runtime in the foreground with no UI.Dashboard tabs
The dashboard is organized into seven tabs, each providing a focused view of a different layer of your project.Overview
Workflow cards with progress pills showing task counts per lifecycle stage, plus per-workflow run and stop controls. This is your primary operational view — you can start and stop individual workflows directly from here without dropping to the command line.Product
Product documents generated when a workflow runs: mission statement, tech stack description, and entity model. These documents are produced by the launch flow and serve as the ground truth for the AI during subsequent runs.Roadmap
A visual task roadmap showing effort estimates, priorities, and dependencies across the entire backlog. Use this tab to assess scope and sequencing before starting a workflow run.Processes
Real-time monitoring of running task-runners — provider, model, task, and elapsed time for each active slot. If you have several parallel workers running, this tab shows you what each one is doing at any given moment.Decisions
Browse the decision log: architectural and design decisions with their status, rationale, and supersession chain. When the AI makes a structural choice during a session, it records a decision entry here so you can review and challenge it later.Workflows
Pipeline-phase view of all tasks with filtering by workflow, phase, and category. Use this tab to drill into a specific workflow’s queue and verify task distribution.Settings
Coding-agent selection, model tier per phase, and permission-mode controls. This is also where coding-agent detection results are displayed — see Coding-agent detection below.Workflow-launch dialog
When you start a workflow, dotbot opens a launch dialog driven by theworkflow.json manifest installed in your project. The manifest declares one or more form modes, and the runtime picks the mode that matches the project state. Each mode shows different fields depending on its visibility flags:
- Prompt field (
show_prompt) — a free-text description of the feature or change to build. Use this for quick, well-understood tasks you can describe in a sentence or two. - File upload (
show_files) — attach a requirements document or specification file. dotbot reads the document and uses it as context for the run. - Interview (
show_interview) — a guided requirements-gathering flow that produces product documents and a task roadmap before generating tasks.
condition that controls when it appears. For example, start-from-prompt only offers its “New Project” mode while .bot/workspace/product/mission.md does not yet exist.
The specific fields and modes available depend on which workflow is installed. For example,
start-from-jira adds Jira and Azure DevOps context to the run, while fast-prompt shows a single prompt field and runs it as one task.Coding-agent detection
The Settings tab detects every coding-agent CLI installed on your system and lets you choose the active one. dotbot supports five providers:
For the active provider you choose a model tier —
fast, balanced, or best (the default) — and a permission mode that controls how the agent handles permission checks during autonomous execution.
You can switch the active provider, model tier, and permission mode from the Settings tab at any time. Changes take effect on the next workflow run.