jcode: a high-performance coding agent harness with semantic memory and collaborative multi-agent swarms

jcode: a high-performance coding agent harness with semantic memory and collaborative multi-agent swarms

What it solves

jcode is a high-performance coding agent harness designed to improve the "skill ceiling" of AI coding assistants. It addresses the resource inefficiency and slow startup times common in existing CLI agents, while providing advanced memory management and collaborative multi-agent workflows.

How it works

  • Resource Optimization: The tool is built for extreme efficiency, boasting significantly lower RAM usage and faster "time to first frame" compared to other coding agents.
  • Semantic Memory: It uses a graph of memories where each turn and response is embedded as a semantic vector. A background "memory sideagent" extracts and consolidates memories, allowing the agent to recall relevant information via cosine similarity checks without consuming excessive tokens.
  • Swarm Intelligence: It supports "swarms" where multiple agents can collaborate in the same repository. The server manages conflicts (e.g., notifying an agent if a file it read was edited by another) and provides messaging channels for agents to coordinate.
  • UI Enhancements: It features a custom terminal implementation (Handterm) for efficient scrolling and a side panel for real-time file updates, diff viewing, and Mermaid diagram rendering.

Who it’s for

Developers who need a highly performant, customizable coding agent that can handle complex, multi-session workflows and collaborative AI-driven development.

Highlights

  • Extreme Performance: Optimized for low RAM overhead and near-instant boot times.
  • Autonomous Swarms: Agents can autonomously spawn teammates to parallelize tasks, turning the lead agent into a coordinator.
  • Semantic Memory Graph: Passive background memory extraction and consolidation for human-like recall.
  • Broad Provider Support: Integrates with major OAuth flows (Claude, OpenAI, Gemini, Copilot) and local runtimes (Ollama, LM Studio, vLLM).
  • Advanced Terminal UI: Includes a high-speed Mermaid renderer and info widgets that utilize negative screen space.

Sources