One panel for every coding agent.
termana is a terminal project launcher for people who live in the terminal. Manage multiple projects, bind a coding agent to each, and launch with one click: the terminal opens, cd's into the project, and starts the agent.
termana is a local-first terminal project launcher that manages multiple projects in one panel and binds a coding agent (Claude Code, Codex, Aider, OpenCode) to each, with a built-in AGENTS.md context editor.
macOS · Windows · Local-first, no account
$ termana launch my-app
→ opening terminal · cd ~/projects/my-app · agent: claude
claude Claude Code v1.x
> Reading AGENTS.md...
> Project context loaded. Ready.
# termana bound: claude · AGENTS.md synced · 1-click launchCapabilities
Why termana
In a multi-agent world, each CLI agent vendor only manages its own config. termana sits in the seams between agents: a project control layer that belongs to no single vendor.
Project cards
Manage every project in one panel. Click the card body to confirm-and-launch, or hit Launch ▸ directly.
Agent binding
Bind an agent per project; termana resolves the launch command. Built-in presets are locked; custom agents are free to add.
One-click launch
Launch = open system terminal, cd into the project, run the bound agent. macOS via Terminal.app, Windows via PowerShell.
Install detection
Detects agents in your actual login shell: fnm / nvm / asdf / volta setups all work. Only real external commands count.
AGENTS.md editor
Edit / preview tabs with markdown rendering. The file is the source of truth: no sync, no drift, no "Generated by" pollution.
CLAUDE.md pointer
Auto-creates an @AGENTS.md pointer when missing and tells you; confirms before converting existing content; never writes through symlinks.
Launch groups
Group projects and Launch all: each project opens its own terminal in one shot.
Local-first
No account, no cloud, no server. Config is a TOML file on disk; move machines or hand off by copying the file.
Compare
termana vs. per-agent tools
Each agent vendor only manages its own config. termana occupies the seams between agents: the one position no single vendor will own.
| Dimension | Per-agent vendor tools | termana |
|---|---|---|
| Multi-project hub | ✗ scattered | ✓ one panel for all |
| Agent-neutral | ✗ vendor-locked | ✓ belongs to no vendor |
| 1-click launch + cd | ✗ manual terminal | ✓ auto open, cd, run agent |
| AGENTS.md editing | partial | ✓ editor + preview + pointer |
| Install detection | ✗ | ✓ login-shell aware |
| Local-first / no account | varies | ✓ pure local TOML |
| Migrate / new machine | ✗ | ✓ copy one file |
Context editor
AGENTS.md, made effortless
AGENTS.md is the agent-neutral context standard: Codex, OpenCode, Cline and new Aider read it directly; Claude Code reads it via a one-line @AGENTS.md pointer. termana is its editor: file is the source of truth, no sync, no drift.
AGENTS.md is the source
termana edits it and writes only AGENTS.md. The file is the truth: no middle layer.
CLAUDE.md is a pointer
A one-time @AGENTS.md pointer, plain text not symlink, safe on Windows and Git.
6 cases handled
All AGENTS.md × CLAUDE.md combinations covered; divergence enters a visible merge block with human/AI confirmation.
Atomic writes
Temp file + atomic replace: no truncated files on failure; CSP locks down unneeded global APIs.
# AGENTS.md
## Project: my-app
Build with Vite + Tauri. Run tests before commit.
- src-tauri/ Rust backend
- src/ TS frontend
# CLAUDE.md (auto-linked)
@AGENTS.mdDownload
Download termana
Pick your platform. termana is local-first: no account, no cloud, no server. Config is a TOML file on disk.
FAQ
Frequently asked questions
Common questions about termana, written so they can be quoted directly.
What is termana?
termana is a local-first terminal project launcher. It manages multiple development projects in one panel, binds a coding agent (Claude Code, Codex, Aider, OpenCode, etc.) to each, and launches a terminal that cds into the project and starts the agent with one click. It also ships a built-in AGENTS.md context editor.
How does termana work with Claude Code?
Once a project is bound to Claude Code, launching opens a terminal, cds into the project, and runs claude. termana also turns CLAUDE.md into a one-line @AGENTS.md pointer, so Claude Code reads the same unified context you maintain in AGENTS.md; no need to keep two files in sync.
Is termana free? Does it need an account or internet?
termana is local-first: no account, no cloud, no server. Config is a TOML file on disk. It is currently free; the license is not yet finalized.
Which platforms and terminals does termana support?
macOS launches via Terminal.app (osascript); Windows launches via PowerShell (CREATE_NEW_CONSOLE, profile loaded). Install detection runs in your login shell, so fnm / nvm / asdf / volta setups are detected correctly.
How do AGENTS.md and CLAUDE.md relate?
AGENTS.md is the agent-neutral context standard: Codex, OpenCode, Cline and new Aimer read it directly; Claude Code reads CLAUDE.md, but a one-line @AGENTS.md pointer in CLAUDE.md imports the canonical file. termana edits AGENTS.md as the source of truth and creates CLAUDE.md as a one-time pointer: two filenames, one body of content, no drift, no sync step.
How is termana different from each agent vendor’s own tooling?
Each agent vendor only manages its own config; termana occupies the seams between agents as a project control layer that belongs to no single vendor. It provides multi-project hub, agent-neutral binding, one-click launch + cd, AGENTS.md editing, and local-first TOML config, things a single-vendor tool will not offer.
How do I install termana?
Download the build for your platform from GitHub Releases (.dmg on macOS, .msi on Windows) and install it. For development, clone the repo and run npm install && npm run tauri dev.
Roadmap
Roadmap
Where termana is heading.
Project registry + agent binding + 1-click launch
- Project cards & CRUD
- Agent binding + launch resolution
- Login-shell install detection
- Configurable agent list
- Launch groups
AGENTS.md context editor
- Edit / preview tabs
- CLAUDE.md @AGENTS.md pointer
- 6-case handling
- Legacy migration + atomic writes
- Divergence detection & alignment
Deeper per-project config
- Model selection
- Permissions
- MCP config binding
Cross-agent observability
- Global agent session state
- Agent handoff + shared context
- Cross-session handoff