Interrupt mid-turn
You don't hand an agent the keys and hope. Warden gives you the levers to stop a running turn, refuse a single tool, cap the spend, and pick up cleanly after a crash.
Read post →
Engineering blog
Architecture deep-dives, multi-provider design notes, and product updates from the team building Warden, a lean, terminal-native LLM coding agent.
You don't hand an agent the keys and hope. Warden gives you the levers to stop a running turn, refuse a single tool, cap the spend, and pick up cleanly after a crash.
Read post →
Warden's terminal interface is now the default. Search the whole transcript, jump to any command, and open any tool call to read exactly what the agent did.
Read post →
Warden gives you one reasoning-effort control, off through max, that maps onto every provider it drives, shows you the thinking tokens it's spending, and respects your session cost cap. Set it per session, per command, or per agent.
Read post →
Warden now drives OpenAI's subscription-tier coding models over a proper OAuth sign-in, right alongside Anthropic and Gemini, so the plan you already pay for can run your agent. The sign-in flow works even on headless WSL boxes.
Read post →
Warden's default interface is a full-screen terminal UI: tool results render as collapsible cards holding the full, untruncated output, file edits render as diffs, and a managed scroll buffer means nothing scrolls off into the void.
Read post →
Warden is the lean, multi-provider coding agent that actually does the work. Alone it beats the closed CLIs; wired into the stack it edits against ground truth and never flies blind.
Read post →
Warden ships the audit trail on token economics. TheAuditor replaces read-and-guess with a query. The pairing isn't a coincidence.
Read post →
A lean, multi-provider, terminal-native LLM coding agent. Single binary. MIT. Pre-Alpha ships soon.
Read post →