Works with the AI Agents you already use
Will It Run
Checks whether a local LLM fits your GPU, Apple Silicon Mac, or unified-memory machine. Uses the free Will It Run MCP and returns a fit verdict, memory need, offload plan, and measured or estimated speed.
Free
Will It Run
Example session with this skill installed
Can my RTX 5060 8GB with 32 GB RAM run Qwen3.6 35B A3B at 96k context with q8_0 KV cache?
- Read your context and instructions
- Compiled the will it run
This is a live result from the MCP on 12 Nov 2026, Eastern Time, against data revision 2026-11-06-d1cc256. Numbers can change as the database updates.
Request: Can my RTX 5060 8GB with 32 GB RAM run Qwen3.6 35B A3B at 96k context with q8_0 KV cache?
The agent called check_fit with hw=rtx-5060, model=qwen3-6-35b-a3b, ram=32, ctx=98304, kv=q8_0 (quant omitted, so the server picked the best quant that fits).
Result: verdict is offload. The best quant that fits is Q4_K_M. It does not fit fully on the 8 GB GPU. Estimated memory need is 24.56 GB (weights 21.8 GB, KV cache 1.07 GB) against 8 GB GPU-usable memory, so about 16.56 GB goes to system RAM. Offload placement is MoE experts-to-RAM, similar to llama.cpp --cpu-moe: about 24% of layers on the GPU, 1.78 GB of non-expert weights on the GPU, and 16.56 GB of experts in RAM. There was no community benchmark or speed estimate for this exact setup. Checker link: https://willitrunllm.com/?hw=rtx-5060&model=qwen3-6-35b-a3b&q=Q4_K_M&ctx=98304&kvq=q8_0&ram=32&rambw=50#checker
Connects securely to your tools. The creator never sees your data.
About this skill
Will It Run tells an agent whether a local LLM will fit on a given GPU, Apple Silicon Mac, or unified-memory machine (Strix Halo, DGX/RTX Spark). It uses the same free, read-only MCP server as willitrunllm.com: no key and no sign-up.
The agent looks up the hardware and model, then returns a verdict (fits, tight, offload, or no), estimated memory needed vs GPU-usable memory, a RAM-offload plan when relevant (MoE experts-to-RAM or layer split), and speed when a community llama.cpp benchmark exists. It can also list every model that fits a machine, and it can check any Hugging Face GGUF file by real file size.
Memory figures are estimates. Speeds marked measured come from community llama.cpp benchmarks with source links; many setups have no measured speed yet. The free server is for personal, non-commercial use. Full docs are at willitrunllm.com/mcp and terms at willitrunllm.com/terms.
This skill does not install a model or write a complete run recipe. It answers the fit question and links back to the same result on the website.
- Needs a live connection to willitrunllm.com. It will not work offline.
- About 120 tool calls per minute per IP, shared with the public API, best-effort.
- The free server is for personal, non-commercial use.
- Memory figures are estimates. Many hardware and model combinations have no measured speed yet.
- The skill answers whether a model fits. It does not install the model or produce a full run configuration.
- Closed API models (hosted ChatGPT, Claude, Gemini) are out of scope.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 5 days ago
- Free with an Agensi account
Requires
Permissions
Network access to willitrunllm.com, to call the read-only MCP at https://willitrunllm.com/api/mcp. No files to write. No secrets. No other hosts.