Edit files, inspect context, and keep notes
File viewing and editing
Section titled “File viewing and editing”Open files from the Files panel, tool cards, project references, Git changes, or pull request files. Nerve displays text, images, and Markdown, can target a line, toggle wrapping, highlight selection matches, and switch Markdown or Mermaid render modes.
Small regular text files inside the project are directly editable. Type in the text view and save with Ctrl/Cmd+S. A modified indicator remains until the draft is saved, and closing one or more modified file tabs asks whether to save, discard, or cancel. Saving is atomic and revision-checked: if another program changed the file after Nerve loaded it, Nerve preserves your draft and reports a conflict instead of overwriting the newer content.
Editing is unavailable for files outside the project, symbolic links, binary files, and text files larger than 1 MiB. Large files are bounded and show truncation state; unsupported binary content cannot be rendered. Use an external editor or approved agent tools when the built-in editor is not suitable.
Context and active agents
Section titled “Context and active agents”The context popover beside the composer shows context usage and the conversation’s token totals, with a button to compact the conversation. Three icon buttons in its header hold the secondary details: hover the clipboard to see session details (project, directory, conversation ID) and click it to copy them, use the gauge to choose this conversation’s compaction profile (Aggressive, Balanced, or Conservative; it inherits the project or user setting until you pick one, and the reset icon returns to it), and hover the document icon to read the conversation’s custom instructions or click it to edit them. Sub-agents appear under Agents in the activity island above the composer; see Agents and delegation.
The composer meter shows current usage as a fraction of the model’s declared context window when usage information is available. It can display ? before the provider reports enough information. Automatic compaction creates an explicit transcript event. The full stored history graph remains available even when the active model context contains a summary.
Notes are project-scoped scratch records in Nerve state. Create and delete them from the Notes dock panel. They are useful for temporary observations that should not become repository instructions or source files.
For durable agent guidance, use AGENTS.md, SYSTEM.md, or a skill in the documented resource locations. Notes are not automatically loaded as agent instructions.
Next steps
Section titled “Next steps”After compaction, context usage is unknown until the next successful model response. Checkpoint “after” and savings figures are estimates of summary plus retained messages, not exact full-request usage. Signed reasoning is included conservatively in retention budgets; a complete newest tool group may exceed the requested retention target. Failed summary generation leaves the existing context unchanged.