Aivo
Run familiar coding agents with your choice of hosted or local model through one CLI.
- Source repo
- yuanchuan/aivo
- Stars
- ★ 191
- Last updated
- 18d ago
- License
- MIT
- Primary language
- Rust
- FA score
- 60/100 · Some gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformCodex · Claude Code
- Cost
- Free software; you pay for model usage
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A developer who alternates between Claude Code and Codex and wants to launch both from one CLI while choosing a model per task.
- Not a fit if
- Developers who need a GUI or IDE extension
- Source review
- 60/100 · Some gaps
What does this agent do, and when should you use it?
Aivo is a terminal tool that launches coding agents including Claude Code, Codex, Gemini, OpenCode, Pi, and Grok while letting the user select the model and provider. The project also ships Aivo Code, its own terminal coding agent with session tools, skills, and MCP server support. Its model boundary spans hosted APIs, AI gateways, and local GGUF files; the examples include Vercel AI Gateway and a Hugging Face model intended for LM Studio. Aivo keeps API keys in encrypted local storage and provides a consolidated view of sessions, logs, and token usage. It is delivered as a command-line application for macOS, Linux, and Windows rather than a documented graphical development environment.
The user runs aivo in a terminal and selects an agent plus an optional provider or model. Aivo then launches the corresponding Claude Code, Codex, Gemini, OpenCode, Pi, or Grok workflow and routes model access through the selected hosted service, AI gateway, or local GGUF model. For additional providers, aivo keys add collects credentials and Aivo manages them in encrypted local storage. Running aivo code invokes the built-in Aivo Code agent, which can use session tools, skills, and MCP servers. Aivo also brings sessions, logs, and token-usage information into one view.
- A developer who alternates between Claude Code and Codex and wants to launch both from one CLI while choosing a model per task.
- A team evaluating hosted models or AI gateways without wanting to abandon the entry points of its existing coding agents.
- A new user who wants a first trial without provisioning an API key, using the built-in
aivo/starterprovider. - A developer with a local GGUF model who wants to use it from the terminal through Aivo Code.
- A multi-agent user who wants sessions, logs, and token usage collected in a unified view.
- A terminal-focused engineer who wants a built-in coding agent with session tools, skills, and MCP server support.
How do you install or deploy this agent?
On macOS or Linux, run curl -fsSL https://getaivo.dev/install.sh | bash. With Homebrew, run brew install yuanchuan/tap/aivo. In Windows PowerShell, run irm https://getaivo.dev/install.ps1 | iex. The supplied material does not state detailed OS-version requirements for the packaged installs; the repository badge lists rustc 1.97+ as its minimum Rust version for source builds.
How do you use this agent?
For a first run without an API key, execute aivo "tell me a short story" or start Claude Code with aivo claude. To access more models, run aivo keys add, supply the relevant provider credential, and then use a command such as aivo claude --model moonshotai/kimi-k2.5. Other documented launch commands are aivo codex, aivo gemini, aivo opencode, aivo pi, and aivo grok. Start the built-in agent with aivo code; documented model examples include aivo code vercel::zai/glm-5.2 for Vercel AI Gateway and aivo code hf:lmstudio-community/Olmo-3-1025-7B-GGUF for a local GGUF model. A one-shot invocation is shown as aivo code -e "今天成都的天气".
What are this agent's strengths and limitations?
- One CLI directly covers multiple named coding agents, including Claude Code, Codex, Gemini, OpenCode, Pi, and Grok.
- Model access spans hosted APIs, AI gateways, and local GGUF files, reducing dependence on one model provider.
- Aivo Code provides an included terminal agent with skills, session tools, and MCP server support.
- It combines encrypted local API-key handling with unified session, log, and token-usage visibility.
- The built-in
aivo/starterprovider enables a first run without an API key.
- Models beyond the starter option generally require users to obtain and configure the relevant provider credentials.
- Only terminal workflows are documented; no desktop UI, web UI, or native IDE extension is described.
- Local GGUF support is stated, but hardware requirements, expected performance, and detailed model compatibility are not documented in the supplied material.
- There is no evidence showing whether every native agent feature survives a model-provider substitution, so important agent/model combinations require validation.
- Source builders need to account for the stated rustc 1.97+ minimum.
How does this agent compare with similar options?
Compared with invoking Claude Code, Codex, Gemini, OpenCode, Pi, or Grok separately, Aivo adds a common launcher, model and provider selection, local credential management, and a unified view of sessions, logs, and token usage. Aivo Code is an additional built-in terminal agent that can work with remote or local models. No quality, latency, or cost benchmarks are provided, so the source does not establish that Aivo Code outperforms any of the standalone agents.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Aivo This agent | 60 · Some gaps | CLIFree + model costs | ★ 191 | 18d ago | Rust | Codex · Claude Code |
| AgenC Core | 76 · Good | CLIFree + model costs | ★ 114 | today | TypeScript | — |
| zerostack | 69 · Some gaps | CLIFree + model costs | ★ 1.7k | 4d ago | Rust | OpenAI API · Claude API |
| ClawCodex | 65 · Some gaps | CLIFree + model costs | ★ 926 | today | TypeScript | ChatGPT · OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
The README describes a unified way to use multiple coding agents, models, and providers. CLAUDE.md says API keys are encrypted in local configuration, and Cargo.toml lists encryption and Linux write-confinement dependencies. This supports moderate scores for least privilege and sensitive-data handling. Deductions: the packet does not explain per-agent authorization boundaries, key lifecycle, or concrete data flows, and shows no confirmation mechanism for external calls, recovery process, or clear maintainer update commitment.
The README, Cargo.toml, and CLAUDE.md are broadly consistent about a Rust CLI, multiple agents, and local key management. CLAUDE.md describes error exit codes, while the CI and deny workflows show planned automated checks. Deductions: evidence about dependency availability and user-facing runtime failure messages is limited; workflow files do not establish check results or actual reliability.
The README covers hosted APIs, gateways, local GGUF models, and several agents, with examples for basic use, model selection, and Aivo Code. This clearly demonstrates multiple scenarios. Deductions: platform or provider limits, precise routing triggers, and environment support boundaries are only partly documented.
The README provides installation options, a quick start, documentation entry point, and MIT license; Cargo.toml states package version and license, and CLAUDE.md describes project structure and development conventions. Deductions: installation notes are brief, examples and FAQs are incomplete, and the supplied material does not include user-facing known limitations or change history. CLAUDE.md's release instructions are contributor guidance and do not establish maintenance responsibility on their own.
A unified CLI lets users choose among coding agents, models, and providers. Built-in Aivo Code, session tools, and token-usage views also indicate concrete product value, and the quick-start commands are understandable. Deductions: the packet does not demonstrate output formats, realized workflow gains, or cost comparisons, so the utility judgment relies mainly on product descriptions.
Some README claims are corroborated by Cargo.toml package details, encryption dependencies, and CI/deny configuration; CLAUDE.md adds architectural and local key-storage descriptions. Deductions: many security and feature claims remain textual, the packet does not tie each claim to implementation or independent evidence, and it does not clearly separate verified facts from inference.
- The installation commands fetch scripts from the project site and execute them through a shell; inspect the script before running it.
- The packet says API keys are encrypted in local configuration but does not show key-protection, backup, or recovery details.
Maintain this project? Add the FARS badge to your README
[](https://followagents.com/en/agents/aivo)<a href="https://followagents.com/en/agents/aivo"><img src="https://followagents.com/badge/aivo.svg" alt="FARS review"></a>FAQ
Do I need an API key for the first run?
aivo/starter provider works on first run. Additional models require the appropriate provider credentials through aivo keys add.Where are provider credentials stored?
Can it run a fully local model?
Does Aivo replace Claude Code or Codex?
What license does the repository use?
Thanks — we use this to improve the page.