Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add zavora-ai/computer-use-mcp --skill blender-agentgit clone --depth 1 https://github.com/zavora-ai/computer-use-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/zavora-ai/computer-use-mcp/blender-agent)<a href="https://agentmods.dev/skills/zavora-ai/computer-use-mcp/blender-agent"><img src="https://agentmods.dev/badge/skills/zavora-ai/computer-use-mcp/blender-agent/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/zavora-ai/computer-use-mcp/blender-agent"><img src="https://agentmods.dev/badge/skills/zavora-ai/computer-use-mcp/blender-agent.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00035 | $0.02293 |
| Opus 5.5 | $0.00014 | $0.00917 |
| Sonnet 5.5 | $0.00007 | $0.00459 |
| Haiku 4.5 | $0.00003 | $0.00229 |
Grade A, and why
blender-agent scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 18d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 173 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Driving Blender with two MCP servers
Blender needs both servers, but they are not interchangeable. Route by capability, not by habit.
Setup gotchas that cost the most time
These were all found by running it, and each produces a misleading error.
Blender's "Allow Online Access" must be on. The addon's register() silently
refuses to start its server when bpy.app.online_access is false, and every
Blender tool then fails with Cannot connect to Blender at localhost:9876, which
reads like the addon is missing. It is not. Enable it once:
bpy.context.preferences.system.use_online_access = True
bpy.context.preferences.addons['bl_ext.user_default.mcp'].preferences.use_autostart = True
bpy.ops.wm.save_userpref()
With both set, a GUI launch has the socket listening in about two seconds — no clicking through Preferences, and no sidebar panel.
The official and community servers share the name blender-mcp. uvx blender-mcp installs the community package from PyPI. The official Blender Lab
server has to come from projects.blender.org/lab/blender_mcp and be invoked by
path. They also default to the same port, 9876, so run only one.
get_screenshot_of_window_as_image needs a small size_limit_in_bytes. At
v1.0.2 it returns an image at 20 KB and 100 KB, and fails above that with
Invalid response from Blender … Unterminated string, which is a framing bug, not
a permissions problem. get_screenshot_of_area_as_image works fine — prefer it,
since one area is what you usually want anyway.
Render paths are sandboxed to a basename. render_thumbnail_to_path and
render_viewport_to_path discard the directory you pass and write into Blender's
per-session temp directory. Read the filepath out of the result rather than
assuming your own path was used.
Never start a heavy render inside execute_blender_code. A Cycles GPU render
launched through the addon's socket starved it and took the whole Blender process
down — the tool returned Empty response from Blender, then Cannot connect, and
the process was gone with the unsaved scene. Blender's addon server answers
synchronously on the main thread, so anything that blocks for tens of seconds is a
hazard.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 18d ago Changed · +15 lines 3c503f44fd9d
- 1mo ago First seen · 158 lines · 35 tokens per session scan A 4dedc592d08f
blender-agent is a skill published in the GitHub repository zavora-ai/computer-use-mcp (86 stars, last pushed yesterday), licensed MIT. It adds 35 tokens to every session and 2,293 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-12.
Other skills, from other repositories
desktop-gui-inspect
Full-desktop automation for targets that have no browser and no API at all — a legacy Java-based NMS client, a vendor's Windows-only configuration utility, a terminal emulator with no scriptable interface. Drives OpenClaw's ClawHub computer-use skill (Xvfb+XFCE virtual desktop, xdotool input automation) to read…
turix-mac
Computer Use Agent (CUA) for macOS automation using TuriX. Use when you need to perform visual tasks on the desktop, such as opening apps, clicking buttons, or navigating UIs that don't have a CLI or API.
agent-orchestration
A guide for coordinating independent AI coding agents such as Claude Code, OpenCode, or Hermes. It explains how to register each agent, check which ones are running, choose one for a task, and record stable information about how to use it.
workflow-design
A method for designing and testing multi-step computer workflows, including their inputs, actions, conditions, and error handling.
release-poster
A fixed process for creating social-media release graphics in horizontal and vertical formats from a supplied template.
mcp-tools
A reference guide for connecting outside services to an AI workflow through MCP, a standard way for AI systems to call external tools. It describes how to identify a server and tool, provide inputs, and save the result for later workflow steps.