Will this local LLM fit this hardware? Auditable, architecture-aware memory estimates (MLA / sliding-window / hybrid-linear / MoE), with documented runtime/OS reserves. No auth, no key. Engine: MIT (github.com/click6067-ship-it/fitllm-engine) · data: CC0 (fitllm.run/data).
Every API here is available over the APIs.io API and to AI agents over MCP.
One button, every client — Claude, Cursor, VS Code and the rest.
https://apis.io/mcp
find_apisBrowse and filter every API in the catalog.get_api_artifactsOne API's artifacts, grouped by type.get_openapiThe primary OpenAPI for this API.find_similar_apisAPIs that look like this one.apis_io_searchSTART HERE — APIs, providers and tags for one query, each with its total.resolveTurn a domain, URL or GitHub org into the provider it belongs to.find_cohortsEvery scored population of providers in the catalog.curl "https://apis.io/api/v1/apis/fitllm-run:fitllm-agent-api-api"curl "https://apis.io/api/v1/apis?limit=25"Calls without X-API-Key need no key. Calls with it need a paid plan — get a key.
Free tier, no form to fill in. Signing in shares your email address with us — we store it to create your key and to recognise you if you sign in with another provider. See our Privacy Policy and Terms.
A second provider on the same verified email joins the account you already have.
{
"openapi": "3.0.3",
"info": {
"title": "FitLLM Agent API",
"version": "1.0.0",
"description": "Will this local LLM fit this hardware? Auditable, architecture-aware memory estimates (MLA / sliding-window / hybrid-linear / MoE), with documented runtime/OS reserves. No auth, no key. Engine: MIT (github.com/click6067-ship-it/fitllm-engine) · data: CC0 (fitllm.run/data).",
"license": {
"name": "MIT",
"url": "https://github.com/click6067-ship-it/fitllm-engine/blob/master/LICENSE"
},
"x-mcp-server": "https://fitllm.run/api/mcp",
"x-mcp-registry": "https://registry.modelcontextprotocol.io/v0.1/servers?search=run.fitllm%2Ffitllm"
},
"servers": [
{
"url": "https://fitllm.run"
}
],
"externalDocs": {
"description": "Canonical agent workflow, status handling, and citation rules",
"url": "https://fitllm.run/agent/answer-with-citation.md"
},
"paths": {
"/api/check": {
"get": {
"operationId": "checkLlmFit",
"summary": "Check whether an LLM fits on a GPU (incl. multi-GPU rigs) or Apple Silicon Mac",
"parameters": [
{
"name": "model",
"in": "query",
"required": true,
"schema": {
"type": "string"
},
"description": "LLM name, fuzzy — e.g. 'gemma 4 31b'"
},
{
"name": "gpu",
"in": "query",
"schema": {
"type": "string"
},
"description": "GPU name, fuzzy. Multi-GPU: join with + (e.g. '5090+3090'). Provide gpu OR ram."
},
{
"name": "ram",
"in": "query",
"schema": {
"type": "integer",
"minimum": 8,
"maximum": 2048
},
"description": "Apple Silicon unified memory (GB)"
},
{
"name": "quant",
"in": "query",
"schema": {
"type": "string"
},
"description": "GPU: Q4_K_M(default)/Q5_K_M/Q6_K/Q8_0/FP16 · Mac: 4|8(default)|16. Any other value is a 400 with the allowed list."
},
{
"name": "ctx",
"in": "query",
"schema": {
"type": "integer",
"minimum": 16
},
"description": "Context tokens (default 8192, capped at the model's max). A non-integer, below 16 or above the model's max context is a 400."
},
{
"name": "kv",
"in": "query",
"schema": {
"type": "integer",
"enum": [
16,
8,
4
]
},
"description": "KV-cache bits (default 16). Any other value is a 400."
},
{
"name": "env",
"in": "query",
"schema": {
"type": "string",
"enum": [
"windows-display",
"linux-display",
"linux-headless"
]
},
"description": "Discrete-GPU OS reserve per card: windows-display (default, 2 GB) · linux-display (1.2 GB) · linux-headless (0.6 GB). Not accepted for Macs or DGX Spark-class unified-memory systems (400)."
},
{
"name": "count",
"in": "query",
"schema": {
"type": "integer",
"minimum": 1,
"maximum": 8
},
"description": "Copies of gpu"
},
{
"name": "format",
"in": "query",
"schema": {
"type": "string",
"enum": [
"json",
"text",
"md"
]
},
"description": "Default json; curl user-agents get text; md returns a citation block for both computed 200 results and numeric-null 422 statuses"
}
],
"responses": {
"200": {
"description": "Fit verdict with full memory breakdown, max context, receipt permalink, and a computed fix when it doesn't fit",
"content": {
"application/json": {
"schema": {
"type": "object",
"properties": {
"verdict": {
"type": "string",
"enum": [
"yes",
"tight",
"no"
]
},
"receipt": {
"type": "string",
"description": "Stable permalink for this verdict"
},
"model": {
"type": "string"
},
"hardware": {
"type": "string"
},
"env": {
"type": "string",
"enum": [
"windows-display",
"linux-display",
"linux-headless"
],
"description": "Present for discrete GPUs — the OS reserve the verdict used"
},
"quant": {
"type": "string"
},
"kvBits": {
"type": "integer"
},
"ctx": {
"type": "integer"
},
"memory": {
"type": "object",
"properties": {
"usedGB": {
"type": "number"
},
"totalGB": {
"type": "number"
},
"freeGB": {
"type": "number"
},
"breakdown": {
"type": "object",
"properties": {
"weightsGB": {
"type": "number"
},
"kvCacheGB": {
"type": "number"
},
"overheadGB": {
"type": "number"
},
"reserveGB": {
"type": "number"
}
}
}
}
},
"maxContext": {
"type": "integer",
"description": "Same as contextLimits.recommendedMax (kept for compatibility)"
},
"contextLimits": {
"type": "object",
"description": "The two context numbers every FitLLM surface uses",
"properties": {
"recommendedMax": {
"type": "integer",
"description": "Longest context that still leaves the headroom (Mac 20% / GPU 5%)"
},
"hardLimit": {
"type": "integer",
"description": "Longest context that fits at all — beyond it the verdict is no"
},
"headroomPct": {
"type": "integer"
}
}
},
"fix": {
"type": "string",
"nullable": true,
"description": "Concrete actionable change when verdict is not 'yes' — relay verbatim"
}
}
}
}
}
},
"409": {
"description": "Ambiguous model or GPU name — the query matched several catalog entries. The API returns the candidate list instead of picking one, because picking would silently answer about a different model than asked. Re-request with one exact name.",
"content": {
"application/json": {
"schema": {
"type": "object",
"properties": {
"error": {
"type": "string"
},
"candidates": {
"type": "array",
"items": {
"type": "string"
},
"description": "Exact catalog names that matched the ambiguous query."
}
}
}
}
}
},
"422": {
"description": "The exact model is identity-verified but its architecture is not modeled. No fit estimate was computed; cite statusUrl and pinned evidence instead.",
"content": {
"application/json": {
"schema": {
"type": "object",
"required": [
"schemaVersion",
"status",
"model",
"verifiedRevision",
"verifiedAt",
"missingComponents",
"verdict",
"memory",
"usedGB",
"maxContext",
"message",
"statusUrl",
"evidence"
],
"properties": {
"schemaVersion": {
"type": "string",
"enum": [
"fitllm.model-status.v1"
]
},
"status": {
"type": "string",
"enum": [
"unsupported_architecture"
]
},
"model": {
"type": "string",
"enum": [
"Qwen/Qwen3.8-Flash-Next"
]
},
"verifiedRevision": {
"type": "string",
"enum": [
"de4b8e4d43b917e7706784d8bb445c9af86a3540"
]
},
"verifiedAt": {
"type": "string",
"format": "date"
},
"missingComponents": {
"type": "array",
"items": {
"type": "string",
"enum": [
"QSA_INDEXER",
"NGRAM_RESIDENCY",
"MTP_RESIDENCY",
"HC_ACTIVATION",
"MULTIMODAL_WORKLOAD"
]
}
},
"verdict": {
"type": "string",
"nullable": true,
"enum": [
null
]
},
"memory": {
"type": "object",
"nullable": true,
"enum": [
null
]
},
"usedGB": {
"type": "number",
"nullable": true,
"enum": [
null
]
},
"maxContext": {
"type": "integer",
"nullable": true,
"enum": [
null
]
},
"message": {
"type": "string"
},
"statusUrl": {
"type": "string",
"format": "uri"
},
"evidence": {
"type": "array",
"items": {
"type": "object",
"required": [
"kind",
"url"
],
"properties": {
"kind": {
"type": "string"
},
"url": {
"type": "string",
"format": "uri"
},
"sha256": {
"type": "string"
}
}
}
}
}
}
}
}
},
"404": {
"description": "Unknown model/gpu — do not substitute a nearby model; request the exact ID"
},
"400": {
"description": "A parameter is outside its allowed values (quant, ctx, kv, env, ram). The error names the allowed values — the API never silently substitutes a default for a value you sent.",
"content": {
"application/json": {
"schema": {
"type": "object",
"properties": {
"error": {
"type": "string"
}
}
}
}
}
}
}
}
},
"/api/badge": {
"get": {
"operationId": "fitBadge",
"summary": "shields.io endpoint badge for a model×hardware verdict (live)",
"parameters": [
{
"name": "model",
"in": "query",
"required": true,
"schema": {
"type": "string"
}
},
{
"name": "gpu",
"in": "query",
"schema": {
"type": "string"
}
},
{
"name": "ram",
"in": "query",
"schema": {
"type": "integer"
}
},
{
"name": "quant",
"in": "query",
"schema": {
"type": "string"
}
},
{
"name": "ctx",
"in": "query",
"schema": {
"type": "integer"
}
},
{
"name": "kv",
"in": "query",
"schema": {
"type": "integer"
}
},
{
"name": "env",
"in": "query",
"schema": {
"type": "string",
"enum": [
"windows-display",
"linux-display",
"linux-headless"
]
},
"description": "Discrete-GPU OS reserve (default windows-display). The badge label names it, e.g. 'RTX 4090 (Linux headless)'."
}
],
"responses": {
"200": {
"description": "shields.io endpoint schema JSON"
}
}
}
}
}
}