AskCodi is no longer simply an IDE plug-in for code completion, unit tests, and documentation. Its current product is a local-first AI engineering-agent platform: describe the outcome in one conversation, and Codi maps the codebase, breaks the project into tasks, and assigns specialist agents to work in parallel.
Current product
The desktop app keeps code and project state on the user's machine and uses isolated Git worktrees so multiple agents can work on separate tasks. Codi handles planning, model routing, progress summaries, and review, escalating product decisions, permissions, or missing context to the user.
Users can connect their own Claude or OpenAI subscriptions, while Codi selects models according to task difficulty. AskCodi also offers an OpenAI-compatible API for routing other clients across multiple model providers.
Who it suits
- Solo developers and small teams that want to parallelize a feature across several engineering tasks
- Users who want code to remain local while multiple models handle planning, implementation, and review
- People who already pay for Claude or OpenAI and want to reuse those subscriptions
Before you start
The core experience currently centers on the desktop app. A macOS Apple Silicon download is available; check the official site for other platforms. Parallel agents can increase model usage and work best with clear requirements, runnable tests, and strict human review.
Pricing
The official site offers a limited trial followed by an annual software subscription. Model subscriptions or API charges are separate, so verify the latest terms before purchasing.