Ships now
v0.5.2 Latest public release: signed and notarized macOS DMGs, Linux packages, Windows installer, guided setup, MCP connection guide, agent health checks, cost model v2, and paid-path plumbing.Local-first AI agent workstation
The control room for AI coding agents.
Tokenburner keeps agent panes, shells, localhost previews, error context, MCP grants, and cost visibility in one native desktop workspace. The agents stay yours. The control surface stays local.
v0.5.2 public release post-v0.5.2 main source line BYO-AI No desktop telemetry
Use Claude Code, Codex CLI, Aider, Gemini CLI, OpenCode, or any local command. Tokenburner does not bundle model usage or upload your project data.
v0.5.2 release loopback mcp paid path beta
$ pnpm test -- --run
TypeError: failed request in WebPane
error badge created locally
window capture + snippet routed
watching vitest
12 passed, 1 failed
mcp doctor: token + grants ok
cost chip: current session
$ setup_doctor
project wizard preview ready
chat pane shell: send disabled
license activates locally
Mainline
post-v0.5.2 main Current source hardens first-run startup policy, New Project flow, update entitlements, Pro agent-slot gates, release checks, and browser-pin modal behavior.Screenshot truth
Window Window capture ships today. Pane and region capture are roadmap items, so the site names them as planned.Model usage
BYO-AI Tokenburner does not bundle Claude, Codex, Aider, Gemini, OpenCode, or any model credits.Default privacy
Local No desktop analytics SDK. PTY output, screenshots, cost data, and project metadata stay on the machine by default.The loop
The product is the feedback loop, not another model.
Multi-agent coding is crowded now. Tokenburner is narrower: it keeps the local surfaces that agents need in one place, then makes context routing fast, visible, and permissioned.
Keep the local work visible.
Run AI CLIs, regular shells, prompt pads, localhost web panes, and disabled provider chat panes in one native split workspace.
Detect failures where they happen.
PTY output is scanned locally for error patterns so broken test runs and stack traces surface without tab hunting.
Send the right context to the right agent.
Route captured window context and terminal snippets to a chosen AI pane. Pane and region capture are planned, not shipped.
Let agents coordinate with explicit grants.
A loopback MCP server lets agents list panes, read permitted context, send messages, spawn panes, and call granted tools with an audit trail. The Settings UI now includes paste-ready MCP config snippets and a connection doctor.
See what each agent is costing.
Per-pane, project, and agent cost attribution keeps multi-agent work from turning into a mystery bill.
Current limit: the public release and current source line still ship window screenshot capture only. Pane and region capture are planned because they are the modes that make screenshot routing feel complete.
Why it exists
Editors, terminals, cloud agents, and CLIs all solve different pieces.
Tokenburner is the local cockpit between them. It is built for the moments where the work is spread across agents, logs, previews, screenshots, and permissions.
For the multi-agent mess
Replace the pile of terminals, tabs, screenshots, copied stack traces, and setup guesses with one workspace built for parallel agent sessions.
For local feedback loops
Error badges and routing shortcuts turn failing output into a directed debugging prompt without sending the whole project away.
For agent control
Agents can coordinate through Tokenburner tools only after explicit local grants. The app now helps generate MCP config snippets instead of making users spelunk dotfiles.
For people who already chose their tools
Keep your editor and your model accounts. Tokenburner is the workstation around them, not another editor or model provider.
Boundaries
The site should be as clear about the no-list as the feature list.
Tokenburner wins by staying narrow: local workstation, explicit control, no hidden cloud layer, no bundled model credits.
Not a cloud agent platform
No hosted shells, no remote sandboxes, no background cloud workers.
Not a code editor
You keep Cursor, Zed, VS Code, JetBrains, Vim, or whatever already works.
Not model billing
You pay AI providers directly. Tokenburner shows local cost estimates; it does not meter inference.
Not fully done
Pane/region capture, budget kill switches, automated MCP config editing, provider-backed chat, and the fully verified self-serve paid path are still planned.
Privacy posture
Local by default is the product boundary, not a setting buried later.
Tokenburner is built around the assumption that terminal output, screenshots, costs, file paths, and project metadata are sensitive.
PTY content stays local
Terminal output is used inside the app for rendering, error detection, and optional local diagnostics.
Screenshots stay local
Screenshot bytes remain on the machine unless the user deliberately shares or routes them.
Cost data stays local
Token and dollar estimates are local app data, not remote analytics. Cost history/export remain Pro-path work, not shipped public telemetry.
MCP is loopback only
The MCP server binds to localhost and every cross-pane action requires a grant.
No telemetry SDK
The desktop app has no analytics package or hidden phone-home path.
Opt-in crash reports only
Sentry is optional and configured to scrub paths, hostnames, and IP addresses.
License keys verify locally
The app can activate minisign-signed license keys without phoning home. Checkout and license delivery run through separate Cloudflare Worker routes.
Current status
Useful now, still honest about the gaps.
The latest public release is v0.5.2. The current source line is post-v0.5.2 main. The release now includes guided setup, New Project Wizard v2, MCP connection guidance, agent health checks, cost model v2, local license activation, paid-path plumbing, and Windows release assets. Mainline work is now focused on first-run hardening, Pro gates, update entitlement, and closing the remaining local-loop gaps.