> ## Documentation Index
> Fetch the complete documentation index at: https://comfyui-mcp.artokun.io/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Changelog

> What's new in comfyui-mcp.

Weekly notes on what shipped in [comfyui-mcp](https://github.com/artokun/comfyui-mcp).

## Posts

<Card title="July 28: a security patch, reconnect self-heal, and OpenRouter's full catalog" href="./2026-07-28">
  **Path-traversal hardening** on the upload tools, a panel session that
  **self-heals** after a full reconnect/reload/workflow-switch instead of getting
  stuck on a dead tab, the **complete OpenRouter model list**, a Windows
  `restart_comfyui` fix, and a **third-party host** protocol guide. *July 28, 2026*
</Card>

<Card title="July 21: rent a GPU from the panel — RunPod connector and cloud LoRA training" href="./2026-07-21">
  A **first-party RunPod connector** deploys and drives a cloud GPU pod from the
  panel and phone with an honest **local⇄pod host switch**, live cost/GPU status,
  and idle auto-stop; the **LoRA trainer** learns to train on rented pods; plus
  `resolve_missing_models` and provider model discoverability. *July 21, 2026*
</Card>

<Card title="July 20: GPT-5.6, the LoRA trainer lands, Blind mode gets mechanical" href="./2026-07-20">
  The **GPT-5.6 family** (Sol / Terra / Luna) with the max/ultra effort scale,
  seven `train_*` tools that train character LoRAs on FLUX.1-dev from chat, a
  Blind toggle that mechanically gates every image-returning tool, and a revived
  Gemini backend. *July 20, 2026*
</Card>

<Card title="July 19: see what I see — viewport tools and one honest error surface" href="./2026-07-19">
  `panel_view_selected` / `panel_view_nodes_in_viewport` let the agent read what
  you're looking at, and **`panel_get_errors`** becomes the single error surface —
  every errored node joined to its cause, closing the "red node, no error message"
  gap. *July 19, 2026*
</Card>

<Card title="July 18: the 405 that wasn't — speaking pip Manager's legacy-UI dialect" href="./2026-07-18">
  Custom-node and model operations no longer 405 against a pip ComfyUI-Manager in
  legacy-UI mode — comfyui-mcp detects the mode and speaks its `queue/batch`
  dialect. Plus the panel's Kimi K3 chip. *July 18, 2026*
</Card>

<Card title="July 17: Kimi K3, the CivitAI browser grows up, panel-owned sessions" href="./2026-07-17">
  Moonshot's **Kimi K3** joins the provider roster; the CivitAI browser gains
  likes, creator filters, and **load-workflow-onto-canvas** (zip wrappers
  included); and panel-owned sessions become the default — conversations survive
  workflow switches. *July 17, 2026*
</Card>

<Card title="July 16: live queue status everywhere, one-tap mobile cancel" href="./2026-07-16">
  The orchestrator broadcasts ComfyUI's live render/queue state to every client
  (powering the mobile queue monitor), `cancel_job` goes mobile, and
  `list_local_models` links models back to their CivitAI pages. *July 16, 2026*
</Card>

<Card title="July 15: flatten workflows, CivitAI creator discovery, desktop-tab mirror" href="./2026-07-15">
  **`panel_flatten_workflow`** resolves Get/Set buses, Reroutes, and
  use-everywhere links in place without moving a single kept node;
  `search_civitai_creators` makes creator → models → download work end-to-end;
  and the phone can mirror a desktop tab's session. *July 15, 2026*
</Card>

<Card title="July 14: the self-updating orchestrator, vision for every backend" href="./2026-07-14">
  The orchestrator updates and restarts itself when every agent is idle — a
  running install never goes stale. Every Ollama-family backend and
  ChatGPT-direct gains inline image delivery with an honest fallback when a
  model is text-only. *July 14, 2026*
</Card>

<Card title="July 12: five new providers, A2UI chat cards, per-workflow sessions" href="./2026-07-12">
  An OAuth engine brings **Grok, Kimi, GLM, ChatGPT-direct, and Copilot**;
  validated interactive A2UI cards render in chat; per-workflow agent sessions,
  a loopback MCP console, credential slots, native CivitAI search, and official
  comfy-cli integration. *July 12, 2026*
</Card>

<Card title="July 9: LM Studio & llama.cpp backends, phone pairing, gemma4 Arena results" href="./2026-07-09">
  LM Studio, llama.cpp, and custom OpenAI-compatible endpoints become
  first-class backends; token-gated phone pairing and the render mailbox land;
  `query_workflow` filters big graphs; and the gemma4 fine-tune ladder posts its
  Arena scores. *July 9, 2026*
</Card>

<Card title="July 8: local LLMs, remote pods, and the RunPod image — 20 releases in one day" href="./2026-07-08">
  The agent goes all-LLM: an **Ollama backend** with a compact tool mode and our
  gemma4 fine-tune ladder as the local default. Plus one-command `connect` to a
  remote ComfyUI over a secure `wss://` bridge, a custom RunPod image, and both
  ComfyUI-Manager API generations. *July 8, 2026*
</Card>

<Card title="June 29: Comfy MCP parity, run-to-node, graph navigation" href="./2026-06-29">
  Five one-call generation tools close the gap with Comfy's official cloud MCP
  (plus a token-gated tunnel connector), `panel_run` learns to render a single
  output branch, and the panel agent gains graph outline/search/subgraph tools
  plus manual-edit awareness. *June 29, 2026*
</Card>

<Card title="June 27: render-wedge watchdog, color analysis, provider onboarding" href="./2026-06-27">
  A self-healing queue/render watchdog (backpressure + stall detection +
  escalating cancel), a new `analyze_color` tool, registry-first custom-node
  installs with a hardened clone fallback, and a wave of panel polish — plus the
  Krea 2 ULTRA V2 re-slice. *June 27, 2026*
</Card>

<Card title="June 26: strip_workflow & slice_workflow — load any expert graph" href="./2026-06-26">
  Two new workflow tools resolve GetNode/SetNode buses + Reroutes into real
  connections and un-chunk rgthree Fast-Groups-Bypass-toggled pipelines — clean,
  runnable wiring without hand-tracing. *June 26, 2026*
</Card>

<Card title="June 25: Multi-provider agent — Claude + ChatGPT" href="./2026-06-25">
  The sidebar panel agent now runs on **either Claude or ChatGPT/Codex** at full
  feature parity — pick a provider, drive your live canvas, load packs in one shot
  with `panel_load_workflow`, and stay cost-aware about paid API nodes via
  `check_workflow_runtime`. *June 25, 2026*
</Card>

<Card title="June 24: Krea 2 Turbo pack" href="./2026-06-24">
  Day-one ComfyUI pack for Krea 2 Turbo — Krea.ai's new 12B open-weight
  text-to-image model. 8-step turbo speed, Qwen3-VL prompting, JSON area
  prompts, one-command install. *June 24, 2026*
</Card>

<Card title="June 19: v0.15.0 — streaming panel, SDK commands, one-click recovery" href="./2026-06-19">
  The Agent Panel gains live-streaming chat, SDK slash commands (`/compact`,
  `/loop`, …), and one-click `/restart` recovery for a wedged agent — with
  pid-reuse-safe process handling. `--channels` is removed; the workflow converter
  got tougher and more packs are render-verified. *June 19, 2026*
</Card>

<Card title="June 17: v0.14.0 — the panel drives itself" href="./2026-06-17">
  The sidebar panel now runs an autonomous background agent on your Claude
  subscription — no API keys, no interactive session needed. *June 17, 2026*
</Card>

<Card title="Week of June 8: v0.9.5 → v0.13.0" href="./2026-06-16">
  Eight releases: channels mode (drive ComfyUI from your own Claude session,
  no API keys), the sidebar panel live on the Comfy Registry, `generate_audio`,
  Civitai MCP pairing, and a wave of community reliability work. *June 16, 2026*
</Card>

<Card title="Week of June 1: v0.8.1 → v0.9.4" href="./2026-06-05">
  Six releases: Comfy Cloud + explicit remote mode, an embedded agent
  sidebar inside ComfyUI, a new `health_check` tool, slimmer default
  installs, and a stack of stability fixes. *June 5, 2026*
</Card>

<Card title="Week of May 25: v0.6.0 → v0.8.0" href="./2026-05-29">
  Four releases in one week: media uploads, hosted docs, model-download
  hardening, custom-node authoring + verification, declarative environment
  setup, server-side image conversion, and cloud storage I/O. *May 29, 2026*
</Card>