GET /api/v1/hosts/machines/{id}

Operation: getMachine.

Machine detail: compatibility findings, current offer/job, observed metrics, drain state (host owner/admin)

Includes telemetry: the latest agent-reported sample plus observed_at, age_s and boot_id, or null when nothing was ever reported. Telemetry is evidence only and never influences admission, qualification, billing or job state. A measurement the host could not take is ABSENT from the object — it is never rendered as 0 and never carried over from an earlier sample, so a consumer must age it via age_s and show unknown when stale.

Endpoint

GET https://amplerun-api.spiry.org/api/v1/hosts/machines/{id}

Authentication

Use the credential and scope required by this operation. For programmatic renter/host calls: Authorization: Bearer $AMPLERUN_API_KEY. A host-agent signature or other specialized proof is required where the description specifies it.

Parameters

Name	Location	Presence	Type	Description
id	path	Required	UUID

Responses

200

Machine detail.

application/json

Field	Type	Presence	Description
accepts_custom_images	boolean	Optional	The host’s opt-in to renter-supplied public images on this machine (default false).
tier	“container” / “inference”	Optional	Tier A (renter containers) or Tier B (managed inference), set at enrollment.
host_mode	object	Optional	D-22 host mode as the last heartbeat reported it; null until one does.
kernel_driver_runtime	string	Optional	Kernel/driver/runtime triple from the manifest, e.g. linux-6.8/nvidia-560.35/cuda-12.6.
recent_jobs	array of object	Optional	The machine’s latest 20 rentals, newest first. ended_at is null while a rental is open.
listing_blockers	array of string	Optional
market_price	object	Optional	D-35 suggested listing price: the market price a renter pays for ONE of this GPU for one hour. basis amplerun = AmpleRun’s own listed offers of this model (at least 3); otherwise basis reference = a hand-curated table of official public list prices, each with its page and the date it was read (upper side, p75); otherwise basis throughput = each reference row of the same class (consumer or datacenter) turned into the price giving this GPU the same tokens per dollar, from memory bandwidth (LLM token generation is memory-bound), and the p75 of those. Renters pay the listed price (the 10% commission comes out of it), so the suggestion is the market price itself. null when neither exists. A suggestion only: the host sets its own price.
auto_qualify	object	Optional	The automatic machine test: the worker queues it while the agent heartbeats, retries failures with backoff, and admits a passed machine with a listed offer. reason is set for retrying and needs_attention.
telemetry	schema	Optional

{
  "type": "object",
  "properties": {
    "accepts_custom_images": {
      "type": "boolean",
      "description": "The host's opt-in to renter-supplied public images on this machine (default false)."
    },
    "tier": {
      "type": "string",
      "enum": [
        "container",
        "inference"
      ],
      "description": "Tier A (renter containers) or Tier B (managed inference), set at enrollment."
    },
    "host_mode": {
      "nullable": true,
      "description": "D-22 host mode as the last heartbeat reported it; null until one does.",
      "type": "object",
      "properties": {
        "mode": {
          "type": "string",
          "nullable": true,
          "enum": [
            "dedicated",
            "idle",
            "colisted"
          ]
        },
        "offerable": {
          "type": "boolean",
          "nullable": true
        },
        "reasons": {
          "type": "array",
          "items": {
            "type": "string"
          }
        },
        "observed_at": {
          "type": "string"
        }
      }
    },
    "kernel_driver_runtime": {
      "type": "string",
      "nullable": true,
      "description": "Kernel/driver/runtime triple from the manifest, e.g. linux-6.8/nvidia-560.35/cuda-12.6."
    },
    "recent_jobs": {
      "type": "array",
      "description": "The machine's latest 20 rentals, newest first. ended_at is null while a rental is open.",
      "items": {
        "type": "object",
        "required": [
          "job_id",
          "state",
          "started_at",
          "ended_at",
          "rate_micro_per_hour"
        ],
        "properties": {
          "job_id": {
            "$ref": "#/components/schemas/UUID"
          },
          "state": {
            "type": "string"
          },
          "started_at": {
            "$ref": "#/components/schemas/Rfc3339Utc"
          },
          "ended_at": {
            "allOf": [
              {
                "$ref": "#/components/schemas/Rfc3339Utc"
              }
            ],
            "nullable": true
          },
          "rate_micro_per_hour": {
            "type": "string"
          },
          "ready": {
            "type": "boolean",
            "description": "False when the host never reached READY (a refused START): nothing ran and nothing was billed."
          },
          "billed_ms": {
            "type": "string",
            "description": "Metered (billed) runtime in ms, from READY."
          },
          "earned_micro": {
            "type": "string",
            "description": "The host's posted share for this rental, micro-USD."
          }
        }
      }
    },
    "listing_blockers": {
      "type": "array",
      "items": {
        "type": "string"
      }
    },
    "market_price": {
      "nullable": true,
      "type": "object",
      "description": "D-35 suggested listing price: the market price a renter pays for ONE of this GPU for one hour. basis amplerun = AmpleRun's own listed offers of this model (at least 3); otherwise basis reference = a hand-curated table of official public list prices, each with its page and the date it was read (upper side, p75); otherwise basis throughput = each reference row of the same class (consumer or datacenter) turned into the price giving this GPU the same tokens per dollar, from memory bandwidth (LLM token generation is memory-bound), and the p75 of those. Renters pay the listed price (the 10% commission comes out of it), so the suggestion is the market price itself. null when neither exists. A suggestion only: the host sets its own price.",
      "required": [
        "gpu_model",
        "basis",
        "median_micro_per_gpu_hour",
        "sample_size",
        "sources"
      ],
      "properties": {
        "gpu_model": {
          "type": "string",
          "description": "Canonical GPU name the prices were matched to."
        },
        "basis": {
          "enum": [
            "amplerun",
            "reference",
            "throughput"
          ]
        },
        "median_micro_per_gpu_hour": {
          "type": "string",
          "description": "Decimal string, micro-USD per GPU per hour."
        },
        "sample_size": {
          "type": "integer",
          "description": "Listings or reference prices behind the median."
        },
        "sources": {
          "type": "array",
          "items": {
            "type": "object",
            "required": [
              "label",
              "url",
              "accessed"
            ],
            "properties": {
              "label": {
                "type": "string"
              },
              "url": {
                "type": "string",
                "description": "Page where the price can be checked."
              },
              "accessed": {
                "type": "string",
                "description": "YYYY-MM-DD the price was read."
              }
            }
          }
        },
        "estimate": {
          "nullable": true,
          "type": "object",
          "description": "ESTIMATED from memory bandwidth, not measured on this machine. null when the GPU's bandwidth is unknown.",
          "required": [
            "tok_s_8b",
            "micro_per_million_tokens",
            "fits_up_to_b_params"
          ],
          "properties": {
            "tok_s_8b": {
              "type": "number",
              "description": "Estimated generation speed on an 8B Q4_K_M model, tokens/s."
            },
            "micro_per_million_tokens": {
              "type": "string",
              "description": "The per-GPU-hour price per 1M generated tokens at that speed, micro-USD."
            },
            "fits_up_to_b_params": {
              "type": "integer",
              "description": "Largest model that fits, billions of parameters at Q4."
            }
          }
        }
      }
    },
    "auto_qualify": {
      "type": "object",
      "description": "The automatic machine test: the worker queues it while the agent heartbeats, retries failures with backoff, and admits a passed machine with a listed offer. reason is set for retrying and needs_attention.",
      "required": [
        "state",
        "reason"
      ],
      "properties": {
        "state": {
          "type": "string",
          "enum": [
            "testing",
            "passed",
            "retrying",
            "needs_attention",
            "waiting"
          ]
        },
        "reason": {
          "type": "string",
          "nullable": true
        }
      }
    },
    "telemetry": {
      "nullable": true,
      "allOf": [
        {
          "$ref": "#/components/schemas/TelemetrySample"
        },
        {
          "type": "object",
          "required": [
            "observed_at",
            "age_s",
            "boot_id"
          ],
          "properties": {
            "observed_at": {
              "$ref": "#/components/schemas/Rfc3339Utc"
            },
            "age_s": {
              "type": "integer",
              "minimum": 0
            },
            "boot_id": {
              "type": "string"
            }
          }
        }
      ]
    }
  }
}

default

Error envelope. 401 unauthenticated, 403/404 forbidden, 409 conflict, 422 invalid input, 429 rate limited (code UNAVAILABLE, retryable:true, no Retry-After header), 503 unavailable money/provider state with retryable:true.

application/json

Field	Type	Presence	Description
error	object	Required

{
  "$ref": "#/components/schemas/ErrorEnvelope"
}

Static request outline

Replace path/query placeholders and provide the required JSON body, headers and credential from the sections above. This is a request outline, not a command that chooses an amount or starts a rental for you.

curl --request GET 'https://amplerun-api.spiry.org/api/v1/hosts/machines/{id}' \
  --header "Authorization: Bearer $AMPLERUN_API_KEY"

Schema definitions · Download OpenAPI · Authentication · Errors and recovery