Activity by day tokens (input + output)
Runs by status
Runs by mode
Runs by route
Tool usage calls
Stop reasons
Usage by model
| Route | Provider | Model | Runs | Input | Output |
|---|
Most-edited files
Recent edits
Activity by workspace
| Workspace | Runs | Tool calls | Input | Output |
|---|
Agent runs click a row to inspect the conversation
| When | Conversation | Mode | Route | Model | Iters | Tools | In | Out | Time | Stop | Status |
|---|
Conversations
| Title | Workspace | Updated |
|---|
Indexed workspaces click a row to list files
| Workspace | Files | Chunks | Last indexed |
|---|
How DevMind works configured in the chat composer
Each message runs in a mode and on a route, both chosen in the chat panel.
Agent reads, edits, runs commands & delegates to subagents · Ask read-only answers · Plan read-only, writes a plan
heavy powerful model for complex work · light fast/cheap model for easy tasks ·
auto routes per messageHeavy route is auto: Claude when
ANTHROPIC_API_KEY is set, otherwise OpenAI gpt-4o. The model picker also powers Tab autocomplete (light) and Cmd+K inline edit (heavy).In Agent mode DevMind can spawn subagents (explore / plan / general-purpose / code-reviewer) that run in parallel, track a todo list, and make surgical edits with
edit_file. Tool usage and subagent spawns are counted on the Overview tab.