fps-tweaks-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| FPS_POOL_DATA | No | Path to a real measurement pool JSON file to run the server against (e.g. /path/to/pool.json). The expected shape is defined in src/pool.ts and validated with zod on load; without it the server uses the synthetic sample pool generated by `npm run generate`. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
Measured data about which Windows tweaks change frame rate, grouped by hardware class. Rewards are normalised to [-1, 1]. Always report the confidence interval and the number of machines alongside any average: a tweak measured on three machines is not a finding.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| list_hardware_classesA | Hardware classes that have measurements, with how many tweaks and machines each one covers. Call this first: every other tool needs a hardware class, and guessing the name will fail. |
| list_tweaksA | Every tweak the dataset knows about, with a human description and whether it needs administrator rights. This is the catalogue only — it carries no measurements; use tweak_stats or recommend for those. |
| tweak_statsA | What was actually measured for one tweak on one hardware class: average reward, number of measurements, number of distinct machines, a 95% confidence interval and a verdict. The verdict is 'no effect measured' whenever the interval still covers zero. |
| recommendA | Tweaks worth trying on a given hardware class, best first. Ranked by the LOWER bound of the confidence interval, not by the average: that prefers a smaller gain that was measured properly over a larger one seen on two machines. Tweaks whose interval covers zero are excluded. |
| simulate_sessionA | Replays the UCB1 policy from the product against this pool and returns the order in which a fresh machine would try tweaks. ASSUMPTION: each measurement is assumed to return exactly the pool average for that tweak, because real per-run rewards are not stored. The order is therefore realistic; the rewards are not samples. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 5 tools
Each tool occupies a clearly distinct slot: discovery (list_hardware_classes), catalogue (list_tweaks), per-item measurement (tweak_stats), ranked advice (recommend), and policy replay (simulate_session). The descriptions explicitly police the boundaries — list_tweaks says it carries no measurements, tweak_stats says it is one tweak on one class — so an agent cannot sensibly pick the wrong one.
All snake_case and consistently noun-oriented (hardware_classes, tweaks, tweak_stats, session), with verbs list_/recommend/simulate_. The only wrinkle is that three names lead with a verb while tweak_stats is a bare noun phrase, but the pattern is still readable and predictable.
Five tools is well-scoped for a read-only dataset exploration surface, and each one earns its place in the discovery → inspect → advise → simulate workflow. Nothing feels like filler or like a redundant wrapper.
The core lifecycle is covered end to end: enumerate classes, enumerate tweaks, get stats for a tweak/class pair, get a ranked recommendation, and replay the policy. The main gap is the reverse lookup — there is no way to ask which hardware classes a given tweak helps on, which an agent would have to assemble by looping tweak_stats.