jcode

jcode is a terminal AI coding assistant for developers who want to choose their model provider and carry useful context between conversations. It supports local and self-hosted model endpoints alongside cloud services.

• Model connections: You can use Ollama, LM Studio or a custom OpenAI-compatible endpoint, as well as Claude, OpenAI, Gemini and GitHub Copilot.
• Persistent memory: The assistant extracts and stores details from conversations, automatically retrieves relevant memories, and checks stored information for staleness and conflicts.
• Past work search: Agents can search earlier sessions and explicitly save or retrieve memories when they need previous work or decisions.

jcode is written in Rust and is open source under the MIT license. Local OpenAI-compatible endpoints don't require an API key. With a cloud backend, model requests go to the selected provider through its supported login or API connection. Local embeddings add memory overhead, and memory use increases as you add sessions.