Orca CLI
Rust coding agent for the terminal and CI built around DeepSeek models that reads code, edits files and runs commands
- 511 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The agent runs shell commands and edits project files
- On macOS, computer use lets the agent operate other applications with your permission
- Requests to DeepSeek go out with your paid key, and every task spends money from the account
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/echovic-orca-agentDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent runs shell commands and edits project files
- On macOS, computer use lets the agent operate other applications with your permission
- Requests to DeepSeek go out with your paid key, and every task spends money from the account
This is third-party code. Review the repository files before installing.
What it does
Orca is a terminal coding agent written in Rust and built around DeepSeek models. You install it through npm or as a prebuilt binary. The `orca` command opens a text interface, and `orca exec` runs a task without it, for example in a script or CI. The agent reads the DeepSeek key from the DEEPSEEK_API_KEY variable. The agent reads and searches code, edits files and runs shell commands. The `--verifier` flag sets your check command, and a task does not count as done until that check succeeds. The suggest, auto-edit, full-auto and plan (read-only) modes decide what the agent does without asking. On first run in a folder Orca asks whether to trust the project and only then loads its settings, skills and agents. Long commands keep running in the background as tasks, and /tasks shows their output. Larger work uses subagents defined in project files, persistent goals through /goal and project memory that grows after successful turns. The `orca mcp add` command connects MCP servers, and editors talk to Orca over the Agent Client Protocol through `orca --mode=acp`.
Who it is for. For developers with a DeepSeek key who want a terminal agent with adjustable autonomy.
Good fit when
- A task has to run from a script or CI without the text interface
- A build or CI watch runs long, and the command should live in the background while you keep talking to the agent
- Work splits well between subagents you described in files inside the project
- An interrupted session needs to continue: `orca --resume` restores the conversation and `orca exec resume SESSION_ID` resumes a headless task
- A long task needs a cap on turns or money set in advance
Not a fit when
- You have no DeepSeek key or need another model provider: Orca works only with DeepSeek
- Project code must not go to an outside model provider: the agent works through the DeepSeek API
- You install the agent on a minimal Linux image without the ca-certificates package: with no certificates no network request goes through
Example request
Run orca exec with a cargo test check and fix the failing test in the parser crate without touching other modulesLimitations
- Usage is billed at DeepSeek's rates. Orca estimates cost from published prices and refreshes them from models.dev when they are more than a day old. The line `refresh = false` in the `[pricing]` section of `~/.orca/config.toml` turns that off.
- Full-auto mode asks for an explicit Full Access confirmation. Tools and subagents that are already running keep their earlier policy until they finish, and a mode change lasts only until the session ends.
- Computer use works only on macOS 14 or newer and needs system permissions.
- Shared ACP sessions through `orca daemon` and `orca attach` work only on Unix.
How to disable. Remove the package with `npm uninstall -g @blade-ai/orca`. If you installed the binary with the script, delete the orca file and the ~/.orca folder that holds settings, session history and project memory.
FAQ
How do I install Orca?
Run `npm install -g @blade-ai/orca` or install the binary with the script from orcaagent.dev. The package is built for macOS, Linux and Windows on ARM64 and x64.
How do I run Orca in CI?
Use `orca exec` with the task text. Pass long or secret task text through stdin so it stays out of the process command line. A budget stop ends Orca with exit code 4, which the pipeline can handle.
Which DeepSeek model is the default?
With no model set, or auto, Orca uses deepseek-flash (DeepSeek-V4.1-Flash). Pick deepseek-v4-pro through /model, the `--model` flag or a model line in config.toml. The reasoning_effort setting takes low, high or max, and max is the default.
How do I cap spending on a long task?
Set the `[budget]` section in config.toml or use `--max-turns`, `--max-tool-calls`, `--max-cost-usd` and `--max-wall-time-secs`. Without them the agent has no preset turn ceiling.
How do I connect an MCP server to Orca?
The `orca mcp add` command adds a server, and the other orca mcp subcommands list and remove servers and handle sign-in. Local stdio servers and remote ones over streamable HTTP or SSE are supported, with OAuth sign-in or a bearer token from the environment.
Discussion
No questions or reviews yet.
Related
CLI toolsDesktop environment for running Codex, Claude Code, OpenCode and Pi in parallel git worktrees, with a mobile companion
A Python terminal coding agent in the manner of Claude Code: reads the repository, edits files, runs commands, with memory, sub-agents and MCP
An MCP server and CLI for code research: search and understand your project and all of GitHub with fewer tokens
An open-source TypeScript terminal coding agent tuned for DeepSeek models and cheap caching
A terminal coding agent tuned for the deepseek-v4 model: deep thinking, reasoning effort control, Agent Skills and MCP
A coding agent for the terminal and desktop: it does not count "done" without tests or a diff and works with DeepSeek, GLM, Kimi, Claude, Codex and Gemini
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| Orca (this entry) | no | 511 | |
| dao-code | no | 786 | also a terminal agent for DeepSeek, built in TypeScript, while Orca is written in Rust and has orca exec for CI |
| DeepTide | no | 1,107 | a DeepSeek coding agent with a macOS app and a window, while Orca lives in the terminal and takes tasks over ACP |
| Codex CLI | no | 131,439 | OpenAI's terminal agent, while Orca uses DeepSeek models |
README in short
The repository page opens with the install and launch commands, then details the terminal interface: @ mentions, the / menu and the approval panel. Next come a feature list and notes on DeepSeek models and pricing. The README is translated into six languages.