Anveesa CLI
Rust terminal wrapper for OpenAI-compatible APIs and local AI coding CLIs
Open Source Rust OpenAI-Compatible View on GitHubOn this page
Overview
Anveesa CLI is a Rust terminal wrapper for AI providers. It gives you one command, anveesa, and lets each provider be configured as an OpenAI-compatible HTTP provider or a local command provider.
Run anveesa for interactive chat, one-shot prompts, stdin review, and provider-specific runs.
Use hosted gateways and model providers that expose chat completions style APIs.
Local command providersSpawn Codex, Claude Code, GitHub Copilot CLI, or custom commands from the same interface.
Workspace-aware safetyContext, planning tools, and guarded write actions help the model work inside your project without silent workspace changes.
Install
Install from the repository with Cargo. The binary name is anveesa.
| Requirement | Purpose |
|---|---|
| Rust and Cargo | Build and install the CLI locally |
| Provider API key | Required for OpenAI-compatible HTTP providers |
| Optional local CLI | Required only for command providers such as Codex, Claude Code, or Copilot |
git clone https://github.com/PandhuWibowo/anveesa-cli.git
cd anveesa-cli
cargo install --path . --force
Quick Start
Initialize a starter config, set a default provider/model, then run Anveesa interactively or with a prompt.
anveesa config init
export SUMOPOD_API_KEY="..."
anveesa config set-provider sumopod
anveesa config set-model "your-sumopod-model"
anveesa
One-shot Prompt
anveesa "write a git commit message"
Review Stdin
git diff | anveesa ask --stdin "review this diff"
Terminal UI
Running anveesa with no prompt opens the current interactive terminal UI. It shows a boxed welcome screen, the active provider/model, the current working directory, command hints, and approval hints when write tools require confirmation.
┌── Anveesa v0.1.0 ───────────────────────────────────────────────────────┐
│ Hello! │ Tips for getting started │
│ │ /clear reset context │
│ .------. │ /exit or /quit to leave │
│ | o o | │ anveesa ask <q> one-shot │
│ | __ | │ │
│ '------' │──────────────────────────────────────│
│ | | │ │
│ sumopod · kimi-k2.6 │ Commands │
│ ~/projects/anveesa-cli │ /clear reset memory │
│ │ /exit quit session │
│ │ y/a approve tools │
└─────────────────────────────────────────────────────────────────────────┘
──────────────────────────────────────────────────────────────────────────
❯
Line editingThe green ❯ prompt supports full line editing and stores input history next to the config as history.
Existing sessions reopen with a Welcome back! header when the provider, model, system prompt, and working directory match. Active memory is stored as session.json.
Before the first token, Anveesa shows elapsed time with rotating tips, then prints a compact assistant header and streams tokens as they arrive.
Clipboard screenshotsOn macOS interactive sessions, a newly copied screenshot is attached to the next OpenAI-compatible request and deduplicated on later turns.
Slash Commands and Pasted Text
Use /clear to reset memory and remove the saved session, and /exit, /quit, or :q to return to the shell. Large pasted input is collapsed into a compact placeholder in TTY mode so compiler output or logs do not flood the prompt area.
Providers
Anveesa supports OpenAI-compatible HTTP providers and terminal command providers. Check the effective provider list at any time with anveesa providers.
OpenAI-Compatible Providers
| Provider | Environment variable |
|---|---|
openai | OPENAI_API_KEY |
sumopod | SUMOPOD_API_KEY |
openrouter | OPENROUTER_API_KEY |
glm, glm-coding | ZAI_API_KEY |
deepseek, gemini, groq, mistral, xai | Provider-specific API key variables |
github-models, together, fireworks, cerebras, sambanova, nvidia | Hosted OpenAI-compatible gateways |
dashscope, qwen, huggingface, vercel-ai-gateway, perplexity | Hosted provider or gateway API keys |
ollama, lm-studio, vllm, litellm, localai | Local base URLs |
Command Providers
anveesa --provider claude-code "summarize this project"
anveesa --provider codex --model "gpt-5.1-codex" "review this repository"
anveesa --provider copilot --model "gpt-5.1" "explain this function"
Tools and Safety
When tools are enabled, OpenAI-compatible providers can inspect the current workspace, show task plans, and perform controlled write or command actions.
List directories, find files, search text, read capped file snippets, and run a basic web lookup.
Write toolsCreate directories, write files, edit files, and run commands with the configured approval policy.
Secret guardrailsObvious secret files such as SSH keys and .env files are blocked from tool reads.
writes:ask confirms each write/run, writes:auto is enabled with --yes, and writes:off disables writes for non-interactive stdin runs by default.
Approval prompts accept y for one action, a for all remaining write/run tools in the current assistant turn, or Enter to deny. Use --yes only when you intentionally want file writes and commands to run without prompts.
Plans and Tool Limits
For multi-step tasks, compatible providers can call set_plan to display a checklist and complete_task as each item finishes. OpenAI-compatible providers can run up to 32 tool rounds per answer by default; set ANVEESA_MAX_TOOL_ROUNDS to raise or lower the cap, up to 256.
File Change Summaries
When an approved tool creates a directory, writes a file, or edits a file, the terminal shows a concise change summary. TTY output includes a green status marker, the shortened path, added/removed line counts, and a capped diff preview.
● Update(src/lib.rs)
└ Added 2 lines, removed 1 line
42 - old value
42 + new value
43 + additional line
Configuration
Anveesa stores provider configuration in a TOML file. Print the active path with anveesa config path, or override it with ANVEESA_CONFIG.
default_provider = "sumopod"
[providers.sumopod]
kind = "openai-compatible"
base_url = "https://ai.sumopod.com/v1"
api_key_env = "SUMOPOD_API_KEY"
default_model = "your-sumopod-model"
[providers.github-models]
base_url = "https://models.github.ai/inference"
api_key_env = "GITHUB_TOKEN"
[providers.github-models.headers]
Accept = "application/vnd.github+json"
X-GitHub-Api-Version = "2026-03-10"
[providers.anthropic-compatible-gateway]
base_url = "https://your-anthropic-compatible-gateway.example/v1"
api_key_env = "ANTHROPIC_API_KEY"
prompt_cache = true
[providers.codex]
kind = "command"
command = "codex"
args = ["exec", "{model_args}", "{prompt}"]
model_args = ["--model", "{model}"]
OpenAI-Compatible Options
| Option | Meaning |
|---|---|
base_url | Provider base URL; Anveesa sends requests to /chat/completions |
api_key_env | Environment variable that supplies the bearer token |
headers | Extra static request headers for gateways that require them |
prompt_cache | Add cache breakpoints to the static system guidance and last history message; Anthropic URLs enable this automatically |
Command Provider Placeholders
| Placeholder | Meaning |
|---|---|
{prompt} | The user prompt |
{model} | The selected model name |
{system} | The optional system instruction |
{model_args} | Expanded model argument list |
{system_args} | Expanded system argument list |
Commands
The CLI exposes prompt, provider, and config commands through one binary.
| Command | Purpose |
|---|---|
anveesa | Open interactive mode for the default provider |
anveesa "prompt" | Send a one-shot prompt |
anveesa ask --stdin "prompt" | Append stdin to the prompt |
anveesa --yes "prompt" | Auto-approve write and command tools for the run |
anveesa providers | List effective providers |
anveesa config init | Create a starter config |
anveesa config path | Print the active config path |
anveesa config set-provider sumopod | Set the default provider |
anveesa config set-model "model" | Set the default model for the active provider |
anveesa config show | Print the effective config |
常見問題
What is Anveesa CLI?
Anveesa CLI is an open-source Rust command, anveesa, that puts OpenAI-compatible APIs, local model servers, Codex, Claude Code and GitHub Copilot CLI behind one prompt. It is MIT licensed.
How do I install Anveesa CLI?
Download a release for macOS or Linux (x86_64 or arm64) from GitHub, or build it with cargo install --path . from a clone of github.com/PandhuWibowo/anveesa-cli. Then run anveesa config init.
Can it change files or run commands on its own?
Not without asking. Every file write, edit, new folder and command waits for your approval, once or for the rest of the turn, and file changes are shown as a diff.
Does it work with local models?
Yes. Any local server with an OpenAI-compatible API works as a provider, alongside hosted APIs and gateways.