DeepChat open-sources a local-first desktop client for multi-model agent workflows
Roger_M_Taylor · x · 2026-07-22
DeepChat packages models, tools, and agents into one local-first desktop app
DeepChat is an open-source, local-first AI agent desktop client designed to reduce friction when working across separate model apps, tool interfaces, and messaging clients.
What it adds
- Multi-model support for cloud providers and local Ollama models.
- Tape & Trace for recoverable session history, including requests, metadata, entries, and token budgets.
- MCP support via Resources, Prompts, and Tools over StreamableHTTP, SSE, or Stdio.
- Skills and ACP so users can install task-specific skills and run ACP-compatible agents from the model selector.
- Remote control for managing bound sessions from Telegram and other clients.
The pitch is to keep long-running agent work in one desktop interface with persistent state and richer tooling.
More from coding & agent
- MCP may solve one of engineering’s biggest productivity drains: context switching — nijfranck · 2026-07-22
- Hugging Face, PyTorch and Red Hat AI announce a Bengaluru meetup — ariG23498 · 2026-07-22
- Grok is being built to solve real engineering problems, not benchmarks — yunta_tsai · 2026-07-22
- A developer built an MCP server and OAuth2 flow from scratch in PHP — elixon · 2026-07-22
- Open-source skill turns Chinese stories into hand-drawn diary videos — dotey · 2026-07-22
- ISO says RLVR can keep the weight spectrum and learn 2.7× faster — UTEXAS · 2026-07-22