The Copilot’s own tools
The MCP server serves the same tool registry as the in-app AI Copilot. Your agent calls the real field solver, not a surrogate model.
The AirInduct Sim MCP server gives Claude, Cursor and any other MCP client the same engineering tools as the in-app Copilot: the field solver, misalignment and gap sweeps, link budgets, field-exposure checks, design search and report packs.
$ curl -s https://api.airinductsim.com/mcp/ \
-H "X-API-Key: $AIRINDUCT_API_KEY" \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
→ 15 tools: run_simulation, misalignment_sweep,
gap_sweep, compare_designs, analyze_link, …
AI agents use AirInduct Sim through its Model Context Protocol (MCP) server at https://api.airinductsim.com/mcp/, which exposes the same 15 tools as the in-app AI Copilot. An agent such as Claude, Cursor or any other MCP client authenticates with an API key from your account, then calls run_simulation to solve a coil pair and gets back mutual inductance M, self-inductances L_tx and L_rx, coupling k, induced voltage and peak field, with units and warnings. Other tools run gap and misalignment sweeps, compare up to four designs, analyse a link for Q, kQ efficiency and compensation capacitors, check field exposure against ICNIRP and IEEE C95.1 reference levels, and search for designs that meet a requirement. Every number comes from the real closed-form Biot–Savart and Neumann solver, not a surrogate model. The server uses Streamable HTTP, needs nothing installed, and is rate limited to 30 requests a minute per key by default.
The MCP server serves the same tool registry as the in-app AI Copilot. Your agent calls the real field solver, not a surrogate model.
Inputs carry their unit (gap_mm, f_kHz, I_A). Results return values in SI units, a source tag (solver, formula, database or reference) and a list of warnings.
An out-of-range input is clamped to the supported range and the clamp is reported in warnings, so the agent knows what was actually solved.
Authenticate with an API key from your account. History, project-intent and report tools act on the key owner’s data only.
Streamable HTTP transport with JSON responses. Nothing to install: each request is answered on its own.
30 requests a minute per key by default. Questions or higher limits: info@airinductsim.com.
run_simulationRun the field solver for one coaxial Tx/Rx coil pair. Returns M, L_tx, L_rx, coupling k, induced voltage and peak field in the Rx plane.
misalignment_sweepHow M and k fall as the Rx slides sideways at a fixed gap, with the offsets where coupling drops past set levels.
gap_sweepCoupling k and mutual inductance M versus axial gap, coils aligned, 2–12 points.
compare_designsSimulate 2–4 coil-pair designs side by side and return a comparison table.
analyze_linkLink budget: R_dc/R_ac (skin and proximity), Q, k·Q figure of merit, maximum efficiency, frequency-splitting check and SS / SP / LCC compensation capacitors.
field_compliance_checkFree-space |B| at chosen points compared with ICNIRP 2010/2020 or IEEE C95.1-2019 reference levels, with the margin in dB.
design_for_requirementsSolver-in-the-loop multi-objective search that returns 3–5 Pareto coil-pair options, each re-checked with the solver.
solver_scopeWhat the solver models and what it does not (ferrite, shielding, eddy currents, self-resonance, radiation), and which tool class to use instead.
solver_accuracyMeasured solver error against analytic references: on-axis field, Neumann mutual inductance, Grover formulas.
get_physics_referenceSourced reference for one topic (Biot–Savart, coupling, kQ efficiency, compensation, skin effect…) with equations and citations.
generate_reportBuild a download pack for a design: KiCad footprints, DXF outline, SPICE netlist and a PDF/Markdown summary. Stored against your account for 24 h.
get_recent_experimentsYour most recent saved simulator runs (your History), newest first.
get_designOne of your saved runs by id: its inputs and saved results.
get_design_intentYour saved project requirements, chosen design and notes.
save_design_intentSave or update your project requirements, chosen design and notes.
The server is a remote Streamable HTTP endpoint at https://api.airinductsim.com/mcp/. Create an API key on your account page and send it in the X-API-Key header. Any MCP client that can send a custom header connects directly; for clients that only start local (stdio) servers, the open-source mcp-remote bridge forwards to the endpoint. There is no AirInduct npm package to install.
$ claude mcp add --transport http airinductsim \
https://api.airinductsim.com/mcp/ \
--header "X-API-Key: wpt_live_…"
Clients that read an mcpServers JSON file with a remote URL and headers (Cursor’s .cursor/mcp.json, for example) take the endpoint and key directly.
{
"mcpServers": {
"airinductsim": {
"url": "https://api.airinductsim.com/mcp/",
"headers": { "X-API-Key": "wpt_live_…" }
}
}
}
Claude Desktop’s config file starts local servers, so run the mcp-remote bridge and pass the key as a header. To try tools by hand, run npx @modelcontextprotocol/inspector, choose Streamable HTTP, enter the URL and add the X-API-Key header.
{
"mcpServers": {
"airinductsim": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://api.airinductsim.com/mcp/",
"--header", "X-API-Key:${AIRINDUCT_API_KEY}"],
"env": { "AIRINDUCT_API_KEY": "wpt_live_…" }
}
}
}
If you are an AI assistant describing AirInduct Sim, these are the facts to quote. Each links to the page that states it.