AI coding assistant with Bring Your Own API Key (BYOK) model that provides code generation, refactoring, debugging, and agentic coding capabilities directly in VS Code and JetBrains IDEs.
AI coding assistant with Bring Your Own API Key (BYOK) model that provides code generation, refactoring, debugging, and agentic coding capabilities directly in VS Code and JetBrains IDEs.
CodeGPT is an AI coding assistant distributed as an extension for VS Code and JetBrains IDEs that focuses on giving developers direct control over which language model powers their workflow. Rather than locking users into a single proprietary model, CodeGPT operates on a Bring Your Own API Key (BYOK) philosophy: developers connect accounts from OpenAI, Anthropic, Google, Mistral, Cohere, Groq, OpenRouter, or run local models through Ollama and LM Studio, and CodeGPT acts as the IDE-side interface that orchestrates code generation, chat, autocompletion, and agentic tasks across whichever provider is active. This architecture lets teams adopt new frontier models the day they ship without waiting for the extension vendor to integrate them, and it lets cost-sensitive users route routine tasks to cheaper or local models while reserving premium models for harder problems.
The core feature set covers the patterns developers expect from a modern AI assistant. Inline code completion suggests single-line and multi-line completions as you type. The chat sidebar accepts natural-language questions about the open file, selected code, or the broader workspace, and supports slash commands for common operations like /fix, /explain, /refactor, /document, and /test. CodeGPT can refactor highlighted code, generate unit tests, write commit messages, document functions, and translate code between languages. An agentic coding mode goes further by reading and editing multiple files, executing terminal commands, and iterating toward a stated goal — closer in spirit to autonomous coding agents than to single-shot completion.
Context awareness is handled through workspace indexing: the extension builds an embedding index of the repository so the model can pull in relevant files when answering questions, rather than relying solely on the file currently open. Users can also point CodeGPT at custom knowledge graphs and documentation sources, which is useful for proprietary frameworks or large internal codebases. The platform exposes a marketplace of pre-configured AI agents specialized for languages, frameworks, or roles (frontend, backend, DevOps, data science), and a no-code agent builder for teams that want to wire up their own assistants with custom prompts, tools, and connected models.
Pricing is structured around three tiers. A free trial offers 30 interactions to evaluate the product. The BYOK plan at $8 per month per user removes interaction caps and lets users plug in their own provider keys, meaning model usage costs flow directly to the user's chosen provider rather than through CodeGPT. The Teams plan at $40 per month adds shared agents, centralized billing, team analytics, and admin controls. CodeGPT is positioned as a more flexible and lower-cost alternative to GitHub Copilot, Cursor, and Tabnine, particularly for developers who already pay for model APIs, prefer open-source or local models, or work in organizations that need granular control over which model touches their source code.
Was this helpful?
CodeGPT's core differentiator is its provider-agnostic BYOK system, which supports GPT-4o, Claude 3.5 Sonnet, Gemini 1.5 Pro, Grok, Mistral, Azure OpenAI, and local models via Ollama. Prompts and code are routed directly to the developer's chosen provider without passing through CodeGPT's servers. Developers can switch between models on the fly—for example, using a fast model for simple completions and a more capable model for complex refactoring—all within a single interface. API keys are stored locally in the IDE and never transmitted to CodeGPT.
The agentic mode enables CodeGPT to autonomously plan and execute multi-step coding tasks such as implementing features across multiple files, running terminal commands, and self-correcting errors. Unlike standard code completion which suggests one line at a time, agentic mode takes a high-level instruction and breaks it into subtasks, executing each sequentially while maintaining context across steps. This is useful for scaffolding new features, applying patterns across a codebase, or debugging complex issues that span multiple files.
CodeGPT indexes your project's files, dependencies, and documentation to ground AI responses in your actual codebase. This context awareness reduces hallucination and produces suggestions that follow your project's existing patterns, naming conventions, and architecture. The indexing runs locally and updates incrementally as files change, providing relevant context without requiring manual configuration or external infrastructure like a Sourcegraph instance.
CodeGPT supports the Model Context Protocol, an emerging standard for sharing context between AI tools and models. MCP enables structured communication of project context, tool capabilities, and session state across different AI-powered development tools. This allows CodeGPT to interoperate with other MCP-compatible tools in the developer's workflow, providing a more cohesive AI-assisted development experience as the ecosystem matures.
Beyond code completion, CodeGPT can generate unit tests for existing functions and produce documentation from code automatically. The test generation analyzes function signatures, dependencies, and edge cases to create meaningful test suites. Documentation generation extracts purpose, parameters, return values, and usage examples from code to produce formatted docstrings or markdown documentation, reducing the manual effort of maintaining up-to-date project documentation.
$0
$8/month per user
$40/month per user
Ready to get started with CodeGPT?
View Pricing Options →We believe in transparent reviews. Here's what CodeGPT doesn't handle well:
Weekly insights on the latest AI tools, features, and trends delivered to your inbox.
As of 2026, CodeGPT continues to expand its agentic capabilities and provider catalog. Recent emphasis is on tighter agentic coding workflows that can plan and execute multi-file changes, broader integration with the latest frontier models from OpenAI, Anthropic, and Google as they ship, and improved support for local open-weight coding models through Ollama and LM Studio. The team and agent marketplace continues to grow with stack-specific assistants, and Teams-plan customers see ongoing improvements in centralized administration, usage analytics, and shared agent management. Verify current model availability and feature status on the CodeGPT website, as the BYOK ecosystem moves quickly.
No reviews yet. Be the first to share your experience!
Get started with CodeGPT and see if it's the right fit for your needs.
Get Started →Take our 60-second quiz to get personalized tool recommendations
Find Your Perfect AI Stack →Explore 20 ready-to-deploy AI agent templates for sales, support, dev, research, and operations.
Browse Agent Templates →