Purpose
Codex Infinity is an autonomous programming agent platform for development teams, running on bare-metal VPS with full root access and no cloud service timeout limits. It is suitable for users who want to use plans such as Claude Max and OpenAI Codex for continuous, long-duration autonomous programming tasks.
Main Capabilities
- Autonomous Programming: Once given a task, you can walk away, and the agent plans, implements, tests, and iterates on its own until the feature is delivered.
- Parallel Agents: Multiple agents can work in parallel on the same codebase, for example handling the backend, writing tests, and code review separately.
- Data and Analytics: Built-in SQL studio, supporting candlestick charts, geographic maps, and time-series charts, allowing you to query databases and visualize results instantly.
- Always-On Monitoring: Always-on agents continuously monitor the repository, detect regressions, security issues, and build failures, and fix them automatically.
- Cloud Environment: Each agent has its own independent bare-metal sandbox, equipped with a full toolchain and GPU access, making it easy to test in isolation before safely merging.
- Infinity Git / GitHub: You can host repositories on Infinity Git or connect GitHub, and browse code, fork projects, and submit PRs on a fully open-source platform.
How to Use
- Choose Harness and Model: The harness refers to the agent loop and tools, and the model is the brain. You can run harnesses such as Codex, Claude Code, Gemini CLI, and Pi, connected to GPT, Claude, Gemini, or open-source models.
- Optional Sub-Agents: You can name sub-agents, fix their harness and model, add prompts, and the main agent can delegate difficult problems to them.
- Token Safety Design: Sub-agent output is written to files, and the main agent only reads the tail, reviews the diff, and redoes the difficult parts itself.
- Retain Control: You can turn sub-agents on or off per run, edit their prompts, or disable the entire feature completely.
- Run Locally: You can run the CLI locally on your own machine and bring your own OpenAI Max plan.
Suitable Scenarios
- Teams or individuals that need 24-hour uninterrupted development, fixes, and monitoring.
- Multi-role parallel development and review on the same codebase.
- Development scenarios that require database queries and visual analysis.
- Users who want to control costs through bring-your-own modelkeys or API keys.
Limitations and Pricing
The platform uses transparent orchestration pricing and supports bring-your-own model keys (OpenAI Max, Anthropic) or using API keys. Different tiers correspond to different numbers of concurrent instances, thinking depth, CI Sentinel, and GPU Worker configurations: the basic tier provides 5 concurrent instances, high thinking depth, Creative Studio, and cloud GPU; higher tiers provide unlimited swarm instances, x-high thinking, always-on CI Sentinel, and usage-based GPU Workers. Specific price figures, free trials, supported regions, or system requirements are not provided in the materials.
