Skip to content

Coding Workspace

Bob Lee edited this page Jul 27, 2026 · 2 revisions

[English] | 简体中文 · Home

BitFun combines an engineering agent with the workspace needed to understand and verify its work. You can begin with a natural-language goal, follow the agent’s reasoning and actions, inspect the actual changes, and take over manually at any point.

From request to verified change

1. Understand

BitFun can map an unfamiliar project, search across the workspace, explain relationships, inspect history, and identify the files or behaviors relevant to a request.

2. Plan

For larger work, use Plan mode to turn a broad goal into reviewable stages. The plan stays visible as the work progresses, so you can redirect before unnecessary changes accumulate.

3. Implement

The agent can edit files, work through terminals, use project tools, and coordinate focused subagents. You can watch the activity in the session or open the same material in the editor.

4. Verify

BitFun can run the project’s own checks, inspect failures, compare expected and observed behavior, and summarize what was verified. Debug mode is available when the root cause—not just the symptom—needs investigation.

5. Review and recover

Inspect changed files and diffs, ask for an AI-assisted review, or run deeper review flows for high-risk work. Session snapshots and rollback controls help you return to an earlier state when an approach does not work out.

Everything around the agent

Files and editor

  • Browse, search, create, rename, move, upload, download, and organize workspace files.
  • Edit code and text with language-aware completion, diagnostics, navigation, formatting, and refactoring support where available.
  • Review side-by-side diffs, Markdown, images, plans, and other common project materials.
  • Use inline AI assistance while keeping the surrounding file visible.

Terminal

  • Run project commands in integrated terminals.
  • Keep multiple terminals and split views for different tasks.
  • Follow foreground work or manage longer-running background processes.
  • Let the agent use the same project context you can see.

Git

  • See the working copy, history, graph, branches, worktrees, and file differences.
  • Stage and unstage changes, create commits, and push or pull.
  • Ask AI to draft a commit message from the actual change.
  • Request focused or deeper reviews before accepting a change.

Browser and visual checks

  • Open a browser alongside the workspace.
  • Let approved browser tools inspect pages, follow interactions, and capture visual evidence.
  • Use Computer Use when validation requires a supported desktop application or visual flow.

Sessions

  • Keep work separated into named sessions.
  • Resume previous work with its conversation and activity history.
  • See files modified by a session.
  • Archive, search, rename, or remove sessions as your workspace grows.

Work locally—or where the project lives

BitFun supports local folders as well as remote workspaces over SSH and container-based environments. The explorer, terminal, agent, search, and Git experience follow the selected target so you do not have to mentally translate between a local UI and a remote project.

Learn about remote workspaces →

Review that scales with the risk

Not every change needs the same ceremony. BitFun supports quick AI review for everyday work and deeper, evidence-oriented review for changes that deserve independent validation. Review-focused agents can remain read-only, examine a specific target, and report actionable findings without silently rewriting the work.

A workspace for collaboration, not autopilot theater

BitFun makes useful autonomy possible, but the work remains inspectable:

  • plans can be reviewed;
  • tool activity is visible;
  • permissions can require approval;
  • edits appear as real file changes;
  • checks and failures remain part of the session;
  • rollback is available when exploration takes a wrong turn.

This makes BitFun useful both when you want an agent to drive and when you want a capable partner beside you.

Next: compare Agent modes, customize your agents, or get started.

Clone this wiki locally