Skip to main content
PromptQuorumBuilt for humans. Structured for AI.
Home/Power Local LLM
Overview & Reference

AIClient2API Review: One Local Proxy for Every AI Protocol

·10 min readBy Hans Kuepper · Founder of PromptQuorum · Discovery engine for open-weight & open-source AI

AIClient2API is a free, open-source, self-hosted proxy that converts the client protocols of Gemini CLI, Codex, Grok, Kiro, and Antigravity into a single, local OpenAI-compatible API. Built by justlovemaki and released under GPL-3.0, it runs via Docker or Node.js, exposes an OpenAI-format endpoint plus a web console, and lets any OpenAI-compatible client talk to providers it would not otherwise be able to reach directly.

Key Takeaways

  • AIClient2API (github.com/justlovemaki/AIClient2API) is a free, open-source, self-hosted API proxy — not a hosted service
  • Built by justlovemaki; multiple near-identical forks exist under other GitHub accounts, but this repository is confirmed as the original, highest-star version
  • GPL-3.0 licensed, per the repository's own license field
  • Converts client protocols for Gemini CLI, Codex, Grok, Kiro, and Antigravity into one OpenAI-compatible API, with bidirectional OpenAI/Claude/Gemini protocol conversion
  • Runs via Docker (recommended) or Node.js/npm; includes a web console for configuration and health monitoring
  • Over 8,700 GitHub stars and 1,380+ forks as of this review, with active releases (v3.5.0 shipped September 2026)

📍 In One Sentence

AIClient2API is a free, open-source (GPL-3.0) self-hosted proxy, built by justlovemaki with over 8,700 GitHub stars, that converts Gemini CLI, Codex, Grok, Kiro, and Antigravity client protocols into a single OpenAI-compatible local API.

💬 In Plain Terms

AIClient2API is a small server you run on your own machine that sits between your AI client tools and the AI providers they talk to. Instead of learning five different provider protocols, you point everything at one local address, and AIClient2API translates the requests behind the scenes into the format each provider actually expects.

📌Note: This review is based on the project's own GitHub repository, README, and release history. It does not claim PromptQuorum has run hands-on load testing or security auditing of AIClient2API's proxy behavior.

Read in:
🇺🇸en🇩🇪de🇫🇷fr🇯🇵ja🇨🇳zh🇪🇸es🇧🇷pt🇸🇦ar🇰🇷ko

AIClient2API (github.com/justlovemaki/AIClient2API) is a free, open-source, self-hosted proxy server that converts the client protocols of Gemini CLI, Codex, Grok, Kiro, and Antigravity into a single OpenAI-compatible API, with over 8,700 GitHub stars. This review is a companion to AIClient2API's entry in the Local LLM Software Directory and covers what it actually does, how to run it, and who it fits.

What Is AIClient2API?

AIClient2API is a free, open-source, self-hosted proxy server that unifies the client protocols of several AI providers and CLI tools behind one local, OpenAI-compatible API. It is built by the GitHub user justlovemaki and distributed under the GPL-3.0 license.

  • Product type: a self-hosted proxy/gateway server, not a chat app, IDE plugin, or hosted SaaS product
  • Creator: justlovemaki (GitHub handle); the repository at github.com/justlovemaki/AIClient2API is confirmed as the original, highest-star version among several near-identical forks under other accounts
  • License: GPL-3.0, a copyleft license — distributing modified versions generally requires releasing the source under the same license
  • Repository created July 20, 2025, per GitHub metadata
  • Scale: over 8,700 GitHub stars, 1,380+ forks, and roughly 5 open issues at review time — a small open-issue count relative to its fork count, suggesting an actively maintained project

What Is AIClient2API's Version History?

AIClient2API has shipped frequent point releases through 2026, with the latest tagged version, v3.5.0, published in September 2026.

  • v3.5.0 (September 2026): latest tagged release at review time
  • v3.4.9.1 and v3.4.8 (September 2026): incremental patch releases in the days before v3.5.0
  • v3.0.0 (May 2026): the project's own release notes describe this as adding deeper AI-assisted integration, including automated skill guides and remote API help endpoints
  • April 2026: full OpenAI image generation and editing API support with protocol conversion, per the project's changelog
  • March 2026: Grok protocol support added, including multimodal input and image/video generation, per the project's changelog

What Does AIClient2API Actually Do?

AIClient2API runs as a local server that accepts standard OpenAI-format API requests and translates them into whatever protocol the target provider actually needs, then translates the response back.

  • Protocol conversion: bidirectional conversion between OpenAI, Claude, and Gemini API formats, using what the project describes as strategy and adapter design patterns
  • Supported providers/protocols: Gemini, Claude (reachable via Kiro or Antigravity client protocols), Grok (via cookie or CLI OAuth), Codex, Kimi, MiniMax, OpenAI, plus custom OpenAI-compatible endpoints, per the project's own documentation
  • Account pool management: polls across multiple accounts per provider, with automatic failover and health checks
  • OAuth integration: built-in OAuth flows for Antigravity, Codex, Grok CLI, and Kiro
  • Load balancing: provider selection with a cooldown on HTTP 429 rate-limit errors and cross-provider fallback chains
  • TLS sidecar: a separate Go-based proxy component the project describes as simulating JA3 TLS fingerprints to work around Cloudflare or 403-style blocks on some provider endpoints
  • Web console: a browser-based UI at localhost:3000 for live configuration, health monitoring, and API testing
  • Logging: full request/response capture, which the project positions as useful for auditing or building training datasets

How Do You Configure and Run It?

AIClient2API is configured through a JSON config file (or the web console) that maps incoming requests to one or more backend provider accounts, then run as a long-lived local server process.

  • Start via Docker (recommended): pull the image, mount a local configs folder, and expose the proxy's ports
  • Start via Node.js: clone the repository, run the bundled install script, then npm start
  • Configure provider credentials and account pools in the mounted configs directory, or through the web console at http://localhost:3000
  • Point any OpenAI-compatible client (a chat app, a coding CLI, a custom script) at the local proxy address instead of the provider's own API endpoint
  • Monitor account health, rate-limit cooldowns, and request logs from the web console in real time
bash
docker run -d -p 3000:3000 -p 8086:8086 -p 1455:1455 -p 56121:56121 \
  -v "your_path/configs:/app/configs" \
  justlikemaki/aiclient-2-api

Platform, Pricing, and Licensing

Platform

What AIClient2API states:
A self-hosted server, cross-platform (Mac, Windows, Linux) via Docker or Node.js — no cloud-hosted version.

Cost

What AIClient2API states:
Free and open source. You may still pay whichever provider account or API key you connect it to; the proxy itself has no fee.

Licensing

What AIClient2API states:
GPL-3.0, a copyleft license, per the repository's license field.

Install method

What AIClient2API states:
Docker (recommended, single docker run command) or Node.js via a bundled install script and npm start.

Verify the current recommended install method and port list on github.com/justlovemaki/AIClient2API before running either command.

How Much Does AIClient2API Cost?

AIClient2API itself is free — it has no license fee, subscription, or usage-based charge. Your actual cost depends entirely on which provider accounts or API keys you connect it to.

  • The proxy software: free, GPL-3.0 open source, no paid tier
  • Provider costs: whatever the underlying account or API key you route through already costs (a paid OpenAI/Claude/Gemini API key, or a subscription tied to a CLI tool like Codex or Kiro)
  • Self-hosting costs: only your own compute — since it is a lightweight Node.js proxy, this is typically negligible even on a small VPS

AIClient2API vs. LiteLLM: What Is the Difference?

AIClient2API and LiteLLM both let you speak one API format and reach many providers, but they target different problems: LiteLLM focuses on production-grade multi-provider routing for standard commercial APIs, while AIClient2API focuses on bridging CLI-subscription and browser-oauth-based AI tools into an OpenAI-compatible format.

Primary goal

AIClient2API:
Bridge CLI/OAuth-based AI tools (Gemini CLI, Codex, Kiro, Antigravity) into one API
LiteLLM:
Unify 100+ standard provider APIs behind an OpenAI-compatible interface

License

AIClient2API:
GPL-3.0 (copyleft)
LiteLLM:
MIT (permissive)

Typical user

AIClient2API:
Developers wanting to reuse CLI-tool subscriptions via a standard API
LiteLLM:
Teams standardizing production LLM traffic across many vendors

Extra features

AIClient2API:
TLS-fingerprint sidecar, OAuth flows for specific CLI tools
LiteLLM:
Budgets, spend tracking, virtual keys, broad ecosystem integrations

These tools solve overlapping but distinct problems — some teams run both, using LiteLLM for standard provider APIs and AIClient2API specifically for CLI-subscription-based providers.

Compare in directory

Who Should Use AIClient2API?

AIClient2API fits developers who already have access to several AI providers through different client tools and want one consistent, local, OpenAI-compatible endpoint to build against.

Use AIClient2API if you

  • Already use multiple AI CLI tools or subscriptions (Gemini CLI, Codex, Kiro, Antigravity, Grok) and want to reach them through one standard API
  • Want a self-hosted proxy you fully control, rather than routing requests through a third-party hosted gateway
  • Need automatic failover across multiple accounts on the same provider to avoid single-account rate limits
  • Want full request/response logging for auditing or building your own training or evaluation datasets
  • Are comfortable running and maintaining a small self-hosted server via Docker or Node.js

Look elsewhere if you

  • Only use one standard provider API key and have no need for protocol conversion or multi-provider routing
  • Want a fully managed, hosted gateway rather than something you self-host and patch yourself
  • Are uncomfortable with the terms-of-service gray area of routing traffic through CLI-tool or browser-OAuth protocols instead of official paid APIs
  • Need enterprise features like team-based budgets, spend dashboards, or SSO — those are more mature in dedicated gateways like LiteLLM

What AIClient2API Is Not Good For

AIClient2API is not a good fit if you need a fully managed gateway, enterprise spend controls, or want to avoid the terms-of-service ambiguity of CLI-protocol bridging.

  • Not a managed or hosted service — you are responsible for running, updating, and securing your own instance
  • Not built primarily around standard, officially licensed provider APIs — several of its integrations route through CLI-tool or OAuth protocols not designed for third-party proxying, which may conflict with a given provider's terms of service
  • Not focused on enterprise governance features (per-team budgets, spend dashboards, SSO) the way dedicated LLM gateways are
  • Not independently security-audited as far as this review could verify — treat it like any self-hosted proxy handling API credentials, and review its account-pool and logging configuration before exposing it beyond your own machine
  • Not backed by a company or funding round this review could verify — treat it as an independently maintained, community-supported project

Common Mistakes When Evaluating AIClient2API

Most confusion about AIClient2API comes from mixing it up with its many forks, or assuming it works like a standard multi-provider gateway.

Mistake 1: Installing a fork instead of the original repository

Several near-identical projects with the same description exist under different GitHub accounts. This review confirms github.com/justlovemaki/AIClient2API as the original, highest-star repository — verify you are pointing at that repository, not a fork, before reporting issues or checking for updates.

Mistake 2: Assuming it only proxies standard, official provider APIs

Several of AIClient2API's integrations route through CLI-tool or browser-OAuth protocols (Kiro, Antigravity, Grok cookie auth) rather than officially supported third-party API access. Check each provider's own terms of service before routing production traffic this way.

Mistake 3: Treating it as a drop-in replacement for a production LLM gateway

AIClient2API does not currently document the team-budget, spend-tracking, or SSO features found in dedicated production gateways like LiteLLM. Evaluate it specifically for CLI-subscription bridging, not as a general enterprise routing layer.

Mistake 4: Skipping the TLS-sidecar and OAuth setup steps

Some of AIClient2API's provider integrations depend on its separate Go-based TLS sidecar or an OAuth flow completing successfully. If a specific provider connection fails, check whether that provider requires the sidecar or an OAuth login step, per the project's own documentation.

Competitors and Alternatives

AIClient2API is most directly compared to LiteLLM; it is also worth evaluating against the OpenAI-compatible servers it can sit in front of, such as vLLM, SGLang, and LocalAI, depending on whether you need a protocol-bridging proxy or the inference server itself.

LiteLLM

Best known for:
MIT-licensed proxy unifying 100+ standard provider APIs with budgets and spend tracking
Link:
LiteLLM review

This list reflects tools commonly evaluated alongside AIClient2API, not an independent PromptQuorum ranking — verify each tool's current feature set before choosing.

Compare in directory

Frequently Asked Questions

What is AIClient2API?

AIClient2API (github.com/justlovemaki/AIClient2API) is a free, open-source (GPL-3.0), self-hosted proxy that converts the client protocols of Gemini CLI, Codex, Grok, Kiro, and Antigravity into a single OpenAI-compatible API.

Is AIClient2API free?

Yes, AIClient2API itself is free and open source under GPL-3.0. You may still pay for whichever underlying provider account or API key you connect it to.

How do I install AIClient2API?

The project's own README recommends Docker: run docker run -d -p 3000:3000 -p 8086:8086 -p 1455:1455 -p 56121:56121 -v "your_path/configs:/app/configs" justlikemaki/aiclient-2-api. A Node.js install script is also available. Verify current instructions on the GitHub repository first.

Which AI providers does AIClient2API support?

Per the project's own documentation: Gemini, Claude (via Kiro or Antigravity), Grok, Codex, Kimi, MiniMax, OpenAI, and custom OpenAI-compatible endpoints.

Who created AIClient2API?

AIClient2API was built by the GitHub user justlovemaki. Several near-identical forks exist under other accounts, but this repository is confirmed as the original, highest-star version.

Is AIClient2API the same as LiteLLM?

No. Both unify multiple providers behind an OpenAI-compatible API, but LiteLLM (MIT-licensed) focuses on standard commercial provider APIs with production features like budgets, while AIClient2API (GPL-3.0) focuses on bridging CLI-tool and OAuth-based AI subscriptions.

Does AIClient2API require a GPU?

No. AIClient2API is a lightweight Node.js protocol-conversion proxy, not an inference engine — it does not run models itself, so it has no GPU requirement.

Is it safe to route traffic through AIClient2API's CLI-protocol integrations?

Several of its integrations route through CLI-tool or browser-OAuth protocols not officially designed for third-party proxying, which may conflict with a given provider's terms of service. Review each provider's own terms before using it for production traffic.

Does AIClient2API support multiple accounts per provider?

Yes, per the project's documentation it supports account-pool management with polling, automatic failover, and health checks across multiple accounts for a given provider.

Has PromptQuorum independently tested AIClient2API's claims?

This review is based on the project's own GitHub repository, README, and release history, rather than hands-on load testing or security auditing by PromptQuorum.

Sources

← Back to Power Local LLM

MCP