All servers / Science / PaceDynamics MCP server
PaceDynamics MCP server
PaceDynamics for AI assistants: vehicle-dynamics and lap-time simulation as an MCP server on your own machine.
Tested by ConnectATDScore 85 out of 100. Tested 2026-11-18 on package version 1.6.1. Package: @pacedyn/mcp · source
Score breakdown
| Runs | 18 / 20 | |
|---|---|---|
| Tool descriptions | 15 / 20 | |
| Package safety | 20 / 20 | |
| Maintenance | 20 / 20 | |
| Adoption | 12 / 20 |
What we saw
- Starts and answers the MCP handshake: yes (1.4 seconds including download)
- Tools offered: 48
- Last published: 2026-11-18
- Weekly downloads (npm): 7,563
- License: SEE LICENSE IN LICENSE · Direct dependencies: 0
Want this server set up for your team?
ConnectATD builds AI automation and wires up MCP servers like this one. Tell us what you are trying to do.
Talk to ConnectATDTools and how they are described
Agents choose tools from these descriptions. Under each name you can see how many of its parameters are explained.
account_statusno parametersWhether this machine is signed in to PaceDynamics, as whom, and whether this copy is the release the app runs. Works without signing in.
account_sign_inno parametersOpens the PaceDynamics sign-in in the person’s browser and waits for them to finish there (up to about a minute and a half; after that it keeps listening and account_status says when it is done). Tell the person a browser tab is opening. Every other tool needs this once per machine.
workspace_statusno parametersWhat this workspace holds: the open project, whether it has unexported edits, when it expires, and any running jobs. The workspace lives in memory only.
project_open3 of 3 parameters describedOpen a PaceDynamics project from a .pace.json on this machine (path), or from its text (document). Replaces whatever this workspace had open.
project_new0 of 1 parameter describedA fresh project: four Formula Student comparison cars, the seeded component libraries, and the bundled track.
project_export2 of 2 parameters describedSave the project as a .pace.json on this machine (path), ready to open in the app; without a path, returns the file’s text instead. A project opened from a file is saved back over it with that same path; any other existing file needs replace: true. The workspace lives in memory, so save before you finish.
project_summaryno parametersCars and the parts they name, the component libraries, tracks, settings and how many results are filed.
results_list0 of 3 parameters describedRuns already filed in this project, newest first — read this before calling anything a baseline. Freshness says whether the car or track has changed since.
results_get0 of 1 parameter describedEverything filed with a run: its events, metrics, points, caveats, the settings it ran under and the inputs it ran with.
results_export_csv0 of 2 parameters describedThe filed results as a CSV table, for a report or a spreadsheet.
results_update0 of 3 parameters describedGive a filed run a note or a folder, so the reason it was run outlives the session.
results_delete0 of 1 parameter describedDrop a filed run from the project.
vehicle_listno parametersEvery car in the project, whether it runs by default, and the library parts it names.
vehicle_get0 of 2 parameters describedThe car’s parameters grouped as the app groups them, with the values a run actually uses (geometry folded in, model readouts applied). Rows marked readout cannot be set — the thing that produces them can.
vehicle_set1 of 2 parameters describedSet parameters by their label. A readout is refused with the reason it is one. Some rows move others with them (the aero balance), and what moved is reported.
vehicle_set_differential0 of 3 parameters describedState the diff on one axle: open, spool, a clutch pack or a Torsen, or null for none (the equal split in a corner, each wheel to its own grip on the envelope). Numbers left out keep the axle’s own when the kind is unchanged, else the kind’s defaults; out-of-range numbers are pulled into range. Only an axle one motor or the engine drives can use a diff: on another the statement is kept, reported as not acting, and acts again once one motor drives that axle. vehicle_get shows each axle’s seat and what is stated.
vehicle_clone0 of 2 parameters describedA copy of a car under a new name — how a concept comparison is built, since points are relative to the field.
vehicle_visibility0 of 2 parameters describedSet whether a car is part of the field a run scores when no ids are named.
vehicle_remove0 of 1 parameter describedDrop a car from the project.
vehicle_assign_part4 of 6 parameters describedPoint a car at a tyre, motor, aero map, battery, brake part, damper part or suspension from the project’s libraries. This is how a concept is varied: sweep the MODEL, not the scalars it supersedes. A tyre goes on both axles unless "axle" says otherwise: a car can run a different tyre at the rear. Fitting or clearing brakes keeps the car braking as it did: the balance bar is seeded from its balance, and clearing writes the balance back. A damper part runs the curves "damperSetup" names, per damper and direction; unnamed ones keep the car’s, else the part’s middle curve. It is every damper the car has: the corner dampers, and the heave and roll elements’ wherever they damp (on a decoupled axle the only ones). "element" fits a part to the heave or the roll dampers alone. A part that states its lengths ("body": extended, stroke, in mm) also sets where those dampers top out and close, over the suspension’s own; readiness says when one does not fit or a wheel has no bump stop.
library_list0 of 1 parameter describedWhat the project holds of each kind, and which cars use it.
library_get0 of 2 parameters describedThe model itself. Large models (an aero map, a fitted tyre) are big — ask for one at a time.
library_import3 of 3 parameters describedAdd a tyre from a .tir file’s text, or an aero map, battery, brake part, damper part, motor or suspension from an exported JSON file’s text. The kind is detected from the content.
track_list3 of 3 parameters describedThe project’s tracks; the bundled courses; and the library catalogue’s tracks, each with its class, event, size, licence and credit — the same list as the app’s Catalogue shelf. Narrow the catalogue by class, event or words. track_select adds any of them.
track_select0 of 3 parameters describedSelect one of the project’s tracks, or add a bundled course by name or a catalogue track by id (track_list has both) and select it. A catalogue track comes with its source and licence, as the app adds it. The session takes the track’s own start: from rest or flying.
track_import7 of 7 parameters describedImport and select a custom track using the app’s importer. Pass file text (or, on this machine, its path): delimited x,y in metres or lat,lon in degrees (at least 3 points in driving order), JSON point arrays or a track object, TUM CSV, EUFS/FSDS tagged cone CSV, or fssim cone YAML. Optional named columns: z (m), width (full width, m), wl/wr (half-widths, m), bank (degrees, + raises left), grip (tyre multiplier, 1 = as tested). Read pacelab://track-import-format for examples and unit variants. Returns the detected topology, banking sign, smoothing and warnings: review these. Adds a new track each time; project_export saves it.
track_geometry1 of 3 parameters describedThe centre-line the track is made of: x/y in metres, and the corridor half-widths where the asset carries them. Enough to draw the course without solving a lap. Decimated to fit; the answer says by how much.
settings_getno parametersThe project’s ruleset, envelope mode, switches, lap-count overrides and numerics.
settings_set1 of 11 parameters describedEnvelope mode, racing line, standing start (the selected track’s own: it comes back with the track), quasi-transient, motor thermal, brake fade, the air temperature, the air at a venue (height and pressure: it sets every car’s density and every engine’s torque), the diffs’ held yaw, lap-count overrides, the ruleset. Overriding a lap count keeps the times real and makes the POINTS meaningless.
run_session2 of 5 parameters describedRun the ruleset’s events and score the field. With no vehicleIds it runs every visible car, which is what you want: Formula Student points are relative to the field scored, so a car run alone takes the maximum in everything.
run_trace4 of 5 parameters describedThe solved path of the last run: distance, speed and the x/y the car drove, point by point. Read from the run that already happened — it does not solve again and is not charged. Decimated to fit a reply; the answer says how far apart the samples are.
readiness0 of 1 parameter describedThe checks the app shows: stand-in models, a course the car cannot steer, an aero platform leaving its data, a pack that cannot hold the rules’ power. Read before trusting a number.
sweep2 of 5 parameters describedRe-run the session across a grid. Axes are a parameter by label, a library part by id (the honest way to price a component), or a switch. Readouts are refused. Above four points it becomes a job.
convergence0 of 3 parameters describedSolve one car’s lap at a ladder of track steps. Quasi-steady discretisation bias is common-mode, so this says how much of an ABSOLUTE time is the grid.
job_listno parametersEverything this workspace has running or finished.
job_status0 of 1 parameter describedState and progress. A finished job’s numbers are read with job_result.
job_result0 of 1 parameter describedThe result of a job that is done. A failed job reports why instead.
job_cancel0 of 1 parameter describedAsk a running job to stop. It stops between two runs, not mid-solve.
study_getno parametersThe optimisation problem the project has open: what may move, what must hold, and what is being minimised. Also lists every study the project keeps.
study_set2 of 5 parameters describedSet the problem from a preset or from a full spec. The presets: front, rear and steering move the wishbone or steering pickups; camber (stand the tyres up), platform (hold the aero balance, needs an aero model) and steerTyre (Ackermann for the tyre) are the questions read at the car’s states. A preset starts a new study, as it does in the app; a spec replaces the open study’s question, or starts one when none is open. It is checked before it is stored, and what it cannot do is reported. A variable is { id, lo, hi }; add around: { minus, plus } to box it around where the car has it (the box then follows the car), or fixed: true to keep it out of the search. step: 1 lands the answer on 1-unit steps from where the car has it; values: [30, 35, 40] on one of those (the box is their span). The search runs continuous and each seed reports its answer onSteps, scored again; study_adopt writes that one while it stays legal. A hardpoint may move along a line or in a plane instead of in x, y and z: hp.<axle>.<point>.line(az,el) is its coordinate along the line through where it is now, and plane(az,el)u with plane(az,el)v its two in the plane with that normal; az is the plan-view angle from x toward y and el the angle up from level, in hundredths of a degree, so line(0,0) runs along x and plane(0,0) is a bulkhead. Box them with around. follows: "<id>" moves a variable with that one, by as much as it moves, and keeps it out of the search: both pivots of a bracket (hp.rear.ucaRear.x follows hp.rear.ucaFront.x, and y and z), the rear as the front (param.arb_rear follows param.arb_front). The two share a unit, the one followed is searched itself, and a point that follows does not also move along a line or in a plane. states: [{ id, name, speedKmh, ayG, axG, weight }] are where the car is when it matters; a target with at: { stateId } is read at one ("*" every state, where the metric combines them), and a state with follows: "slowCorner" | "fastCorner" | "braking" | "traction" | "topSpeed" has its numbers read off the car; one with lap: "<name>" came from a lap (statesFromLap) and keeps its numbers. envelopeFollows: true reads the travel and roll the linkage targets cover off the states’ poses. A preset starts on the car’s five.
study_probe0 of 1 parameter describedMeasures one evaluation and reports the variables that move nothing — before anybody waits for a search. A dead variable is a wiring question, not a result.
study_run1 of 5 parameters describedRun the optimiser from several scattered starts — the Studies tool’s own runs 1–3, so a run here answers what the app answers. Returns a job id: this is minutes, not seconds. job_result carries every seed, how far apart they landed, the edits the best one implies, and asBuilt: each target and hold read with every point it moves built ± tolerance mm off.
study_adopt1 of 3 parameters describedApply a finished search’s answer: geometry to the suspension the car runs, or to a copy of it; the car’s own settings to the car; brake hardware to a fork of the car’s brake part, which the car is then pointed at. This changes the project; export it to keep the change.
garage_compare4 of 5 parameters describedThe Garage’s table: every row of each car’s sheet — Parts, Setup, Characteristics at an operating point, Results of the last run_session — with each value’s difference from the Reference. Run the session first for Results. differencesOnly (the default) keeps the rows the cars differ in.
report_render2 of 3 parameters describedOne self-contained HTML file: a report layout applied to the cars, page for page and number for number as the app’s Print and Download. Layouts: "spec", "comparison", "setup", "review", or one of the project’s own by id or name. Run the session first for Results and the charts.
log_open5 of 5 parameters describedRead a logged session on this machine, as the app’s Calibration tool reads it: the logger, the time and distance, which channels it maps and which it lacks, a channel that died part of the way and when, the laps, and the questions it can answer (a track, the grip used, each fit). Works with no project open; the parameters need one (log_compare).
log_compare5 of 5 parameters describedEvery parameter of the car a logged session measured — aero, resistance, mass, tyre, brakes, powertrain, battery, steering — beside the car’s own value, with one standard error, how far off and the verdict, as the app’s Compare step. Read against the car the session ran as when the log names an issued setup’s code, else the active car (or vehicleId) as it stands. Needs the car’s project open.
log_report8 of 8 parameters describedOne self-contained HTML file of two pages, page for page as the app’s Report step: the run and its setup, data health, laps, the road coloured by speed, the grip used inside the car’s envelope, every parameter against the car with its error bar, energy, and what changed since the session before (previous). Pass out, a path ending in .html, and it is written there — it runs to a few hundred kB; without out, the reply carries it. Needs the car’s project open.
Tested in a sandbox with an empty environment: no API keys, a throwaway folder, nothing from a real account. A passing test shows the server runs and responds. It is not a security audit.