
CodeGPT AI Agents are autonomous coding agents that plan, refactor, and execute tasks across entire codebases.
A bring-your-own-key architecture accepts more than fifteen model providers plus local runtimes, removing single-vendor lock-in and letting teams control model spend per task. Deterministic dependency mapping gives agents full-repository context, with planning steps that can be reviewed before execution.
Self-hosted and on-premise deployment, SOC 2 certification, and IDE-native integration position the offering for regulated and enterprise environments. The company reports more than two million developers across over 180 countries.
CodeGPT AI Agents is the agentic coding offering of CodeGPT, a platform operated by Codegpt LLC and also known as CodeGPT by Judini. The provider was founded in early 2023, and the CodeGPT coding assistant launched in March 2023, distributed as a Visual Studio Code extension first published in December 2022.
The publisher's release notes introduce CodeGPT AI Agents in extension version 2.1.20 alongside the CodeGPT Plus login. That entry is absent from the listing captured in June 2023, which then ended at version 2.1.9, and is present by October 2023, when the listing was retitled Code GPT: Chat & AI Agents.
CodeGPT prices the offering as a freemium ladder: a free tier allows ten daily interactions on economy models plus a one-time $1.50 welcome credit, while paid plans cost $9 per month for individuals and $27 per seat monthly for teams when billed annually.
Usage is anchored on bring-your-own-key access, so developers with existing provider accounts consume no platform tokens, with credit top-ups for premium models. Enterprise and AI-first services engagements are custom-quoted and add SSO, self-hosted deployment, and implementation support.
CodeGPT AI Agents is the agentic coding offering from CodeGPT: autonomous agents that analyze project structure, plan multi-step changes, and edit files across an entire repository. It ships as a Visual Studio Code extension with JetBrains support and runs in planning, refactoring, and debugging modes.
The framework also targets business automation through internal copilots, retrieval-augmented systems built on company knowledge, and cloud or self-hosted workflow automation. Model access is provider-neutral across OpenAI, Anthropic, Google, xAI, Meta, Mistral, and local runtimes such as Ollama.