[tooling] · · 2 min read
Termic launches open-source CLI agent manager that runs the real binaries, not wrappers
Termic spawns your existing AI coding CLIs directly in PTYs, passes through their auth, and avoids the new SDK credit metering — all under an AGPL license.
By ByteBulletin Editors · Editorial Team
A new open-source tool called Termic is taking a novel approach to managing AI coding agents: instead of wrapping vendor CLIs with its own SDK or proxy layer, it spawns the actual CLI binaries (Claude, Codex, Gemini, Copilot, Grok, OpenCode) in pseudo-terminals and inherits their existing authentication. The pitch is that when Anthropic, Google, or OpenAI ship a feature in their native CLI, Termic users get it the same day — no waiting for a wrapper to catch up.
The application, free and AGPL-3.0 licensed, is built with a Rust backend and a WKWebView frontend, coming in at roughly 10 MB versus an Electron equivalent. It supports adding custom CLI tools like Aider or any command that runs in a PTY in about 30 seconds. For teams, per-repo configuration lives in a .termic.yaml file, so setup is shared across clones.
One notable detail is the billing angle: Anthropic's June 15, 2026 Agent SDK credit change applies to the SDK path (and claude -p, GitHub Actions, and third-party apps that authenticate through the SDK). Because Termic spawns the interactive claude CLI directly — the same binary you'd run in a terminal — it stays on the standard Pro / Max subscription, outside the SDK credit meter. The project's messaging leans hard on this: "No per-token markup, no hidden proxy, no custom agent loop between you and the vendor CLI."
Feature-wise, Termic offers tabbed agents, a built-in editor and diff viewer, inline code review comments that batch into a single message, and a "Spotlight" system that mirrors a task's changes into the main checkout as a detached HEAD (so no stray commits land on your branch). It also includes an optional sandbox using macOS Seatbelt and an in-process Rust CONNECT proxy with a hostname allowlist, default-deny on file system access to ~/.ssh, ~/.aws, Keychains, and other sensitive areas.
Under the hood, the editor is CodeMirror (~150 KB), the terminal renderer is a WebGL-based one that avoids row gaps in TUI apps, and the PTY handling uses the same crate as Wezterm. State is managed with plain React hooks — no reducers or thunks. The project is available for macOS and Linux now, with Windows support promised soon.
SOURCES
SHARE
RELATED

[tooling] ·
Whetstone: 20 battle-tested Claude Code skills distilled from real failures
A new open-source plugin turns hard-won lessons from real coding incidents into self-contained skill packs that make AI agents fail loudly instead of silently passing.

[tooling] ·
Anthropic turns Claude Code's auto mode on by default
Claude Code will soon run in auto mode by default, skipping approval prompts unless an action looks irreversible or destructive.

[tooling] ·
Repo Reality Check: A Chrome Extension That Flags Suspicious GitHub Stars
A new browser extension scores GitHub repositories for star anomalies, bus factor, and maintenance health before you commit to a project.
