{"$schema":"https://wellknown.network/schemas/agent-record-v1.json","schemaVersion":"1","id":"ag_k85e4azk9g42","handle":"ecocompute-should-i-quantize-estimator","url":"https://wellknown.network/agents/ecocompute-should-i-quantize-estimator","links":{"self":"https://wellknown.network/agents/ecocompute-should-i-quantize-estimator/record.json","html":"https://wellknown.network/agents/ecocompute-should-i-quantize-estimator","markdown":"https://wellknown.network/agents/ecocompute-should-i-quantize-estimator/record.md","api":"https://wellknown.network/api/v1/agents/ecocompute-should-i-quantize-estimator","status":"https://wellknown.network/api/v1/agents/ecocompute-should-i-quantize-estimator/status","claim":"https://wellknown.network/agents/ecocompute-should-i-quantize-estimator/claim","claimApi":"https://wellknown.network/api/v1/claims","claimDescriptor":"https://wellknown.network/agents/ecocompute-should-i-quantize-estimator/claim.json","badge":"https://wellknown.network/agents/ecocompute-should-i-quantize-estimator/badge.svg","openapi":"https://wellknown.network/openapi.json","history":"https://wellknown.network/api/v1/agents/ecocompute-should-i-quantize-estimator/history","tools":"https://wellknown.network/api/v1/agents/ecocompute-should-i-quantize-estimator/tools"},"ard":{"identifier":"urn:air:ecocompute-estimator.zhanghongping1982.workers.dev:server:ecocompute-should-i-quantize-estimator","type":"application/mcp-server-card+json"},"kind":"mcp_server","declared":{"name":"EcoCompute Should-I-Quantize Estimator","summary":"Energy impact of LLM quantization, calibrated to measured GPU data. Every answer states its basis.","description":"Energy impact of LLM quantization, calibrated to measured GPU data. Every answer states its basis.","publisher":{"name":"hongping-zh","url":null},"homepage":"https://quantenergy.tech","repository":"https://github.com/hongping-zh/quant-energy","version":"1.0.0","license":null,"protocols":["mcp"],"tags":[],"pricing":null,"endpoints":[{"url":"https://ecocompute-estimator.zhanghongping1982.workers.dev/mcp","type":"mcp_streamable_http","auth":null,"probeable":true}],"skills":null,"tools":null,"extra":{"updatedAt":"2026-10-24T03:15:02.110319Z","publishedAt":"2026-10-24T03:15:02.110319Z","registryName":"io.github.hongping-zh/ecocompute-estimator"},"attribution":{"kind":"mcp_registry","name":"mcp_registry","repoUrl":"mcp_registry","summary":"mcp_registry","version":"mcp_registry","description":"mcp_registry","homepageUrl":"mcp_registry","publisherName":"mcp_registry"}},"derived":{"capabilities":[{"slug":"commerce.pricing","name":"Pricing & Quotes","confidence":0.525,"provenance":"derived"}],"categories":["commerce"],"language":"en"},"observed":{"status":"live","statusReason":"Responded 8h ago.","lastOkAt":"2026-11-06T06:28:07.870Z","lastProbedAt":"2026-11-06T06:28:07.870Z","statusComputedAt":"2026-11-06T06:31:16.150Z","reliability30d":{"probes":34,"successRate":1,"p50Ms":100,"basis":"service","measures":{"availability":"availability","latency":"response time","tools":"tool surface observed","summary":"Checks reached the service itself."},"checks":{"total":34,"ok":34,"authBoundaryOk":0,"serviceOk":34,"note":"Counted from the observation rows for the window, checks of the server only (HTTP, A2A card, MCP initialize). ok = authBoundaryOk + serviceOk. `probes` is the sum of daily rollups and includes registry checks, so it can differ from `total`."}},"latestObservations":[{"at":"2026-11-06T06:28:07.870Z","kind":"mcp_initialize","ok":true,"httpStatus":200,"latencyMs":91,"error":null,"detail":{"tools":[{"name":"estimate_quantization_energy","description":"Estimate the task-level energy change (ΔE% vs FP16) of weight-only quantization (NF4 or INT8) for a model of a given size on a GPU architecture class, from meas"},{"name":"get_quantization_crossover","description":"Get the model-size crossover for NF4/INT8 per architecture: below it quantization costs energy, above it saves. Crossovers belong to a card + software stack, no"},{"name":"list_data_coverage","description":"List what is actually measured (GPU architecture × precision × size range, NVML direct), what is borrowed/estimated, and where the gaps are. Use this before tru"},{"name":"search_measurements","description":"Search the measured anchors by model name, GPU name, architecture or size. Returns ΔE% vs FP16 per anchor with run counts — the raw evidence behind the curves."}],"toolCount":4,"toolsHash":"a5d211e858352ff66d46c39c0aea9033166276de0930935e02771d55682bb229","serverName":"ecocompute-estimator","capabilities":["tools"],"serverVersion":"1.1.0","protocolVersion":"2025-06-18"}},{"at":"2026-11-05T21:30:45.007Z","kind":"mcp_initialize","ok":true,"httpStatus":200,"latencyMs":81,"error":null,"detail":{"tools":[{"name":"estimate_quantization_energy","description":"Estimate the task-level energy change (ΔE% vs FP16) of weight-only quantization (NF4 or INT8) for a model of a given size on a GPU architecture class, from meas"},{"name":"get_quantization_crossover","description":"Get the model-size crossover for NF4/INT8 per architecture: below it quantization costs energy, above it saves. Crossovers belong to a card + software stack, no"},{"name":"list_data_coverage","description":"List what is actually measured (GPU architecture × precision × size range, NVML direct), what is borrowed/estimated, and where the gaps are. Use this before tru"},{"name":"search_measurements","description":"Search the measured anchors by model name, GPU name, architecture or size. Returns ΔE% vs FP16 per anchor with run counts — the raw evidence behind the curves."}],"toolCount":4,"toolsHash":"a5d211e858352ff66d46c39c0aea9033166276de0930935e02771d55682bb229","serverName":"ecocompute-estimator","capabilities":["tools"],"serverVersion":"1.1.0","protocolVersion":"2025-06-18"}},{"at":"2026-11-05T12:24:00.733Z","kind":"mcp_initialize","ok":true,"httpStatus":200,"latencyMs":88,"error":null,"detail":{"tools":[{"name":"estimate_quantization_energy","description":"Estimate the task-level energy change (ΔE% vs FP16) of weight-only quantization (NF4 or INT8) for a model of a given size on a GPU architecture class, from meas"},{"name":"get_quantization_crossover","description":"Get the model-size crossover for NF4/INT8 per architecture: below it quantization costs energy, above it saves. Crossovers belong to a card + software stack, no"},{"name":"list_data_coverage","description":"List what is actually measured (GPU architecture × precision × size range, NVML direct), what is borrowed/estimated, and where the gaps are. Use this before tru"},{"name":"search_measurements","description":"Search the measured anchors by model name, GPU name, architecture or size. Returns ΔE% vs FP16 per anchor with run counts — the raw evidence behind the curves."}],"toolCount":4,"toolsHash":"a5d211e858352ff66d46c39c0aea9033166276de0930935e02771d55682bb229","serverName":"ecocompute-estimator","capabilities":["tools"],"serverVersion":"1.1.0","protocolVersion":"2025-06-18"}},{"at":"2026-11-05T04:24:15.740Z","kind":"mcp_initialize","ok":true,"httpStatus":200,"latencyMs":139,"error":null,"detail":{"tools":[{"name":"estimate_quantization_energy","description":"Estimate the task-level energy change (ΔE% vs FP16) of weight-only quantization (NF4 or INT8) for a model of a given size on a GPU architecture class, from meas"},{"name":"get_quantization_crossover","description":"Get the model-size crossover for NF4/INT8 per architecture: below it quantization costs energy, above it saves. Crossovers belong to a card + software stack, no"},{"name":"list_data_coverage","description":"List what is actually measured (GPU architecture × precision × size range, NVML direct), what is borrowed/estimated, and where the gaps are. Use this before tru"},{"name":"search_measurements","description":"Search the measured anchors by model name, GPU name, architecture or size. Returns ΔE% vs FP16 per anchor with run counts — the raw evidence behind the curves."}],"toolCount":4,"toolsHash":"a5d211e858352ff66d46c39c0aea9033166276de0930935e02771d55682bb229","serverName":"ecocompute-estimator","capabilities":["tools"],"serverVersion":"1.1.0","protocolVersion":"2025-06-18"}},{"at":"2026-11-04T20:28:20.715Z","kind":"mcp_initialize","ok":true,"httpStatus":200,"latencyMs":191,"error":null,"detail":{"tools":[{"name":"estimate_quantization_energy","description":"Estimate the task-level energy change (ΔE% vs FP16) of weight-only quantization (NF4 or INT8) for a model of a given size on a GPU architecture class, from meas"},{"name":"get_quantization_crossover","description":"Get the model-size crossover for NF4/INT8 per architecture: below it quantization costs energy, above it saves. Crossovers belong to a card + software stack, no"},{"name":"list_data_coverage","description":"List what is actually measured (GPU architecture × precision × size range, NVML direct), what is borrowed/estimated, and where the gaps are. Use this before tru"},{"name":"search_measurements","description":"Search the measured anchors by model name, GPU name, architecture or size. Returns ΔE% vs FP16 per anchor with run counts — the raw evidence behind the curves."}],"toolCount":4,"toolsHash":"a5d211e858352ff66d46c39c0aea9033166276de0930935e02771d55682bb229","serverName":"ecocompute-estimator","capabilities":["tools"],"serverVersion":"1.1.0","protocolVersion":"2025-06-18"}},{"at":"2026-11-04T12:23:18.033Z","kind":"mcp_initialize","ok":true,"httpStatus":200,"latencyMs":60,"error":null,"detail":{"tools":[{"name":"estimate_quantization_energy","description":"Estimate the task-level energy change (ΔE% vs FP16) of weight-only quantization (NF4 or INT8) for a model of a given size on a GPU architecture class, from meas"},{"name":"get_quantization_crossover","description":"Get the model-size crossover for NF4/INT8 per architecture: below it quantization costs energy, above it saves. Crossovers belong to a card + software stack, no"},{"name":"list_data_coverage","description":"List what is actually measured (GPU architecture × precision × size range, NVML direct), what is borrowed/estimated, and where the gaps are. Use this before tru"},{"name":"search_measurements","description":"Search the measured anchors by model name, GPU name, architecture or size. Returns ΔE% vs FP16 per anchor with run counts — the raw evidence behind the curves."}],"toolCount":4,"toolsHash":"a5d211e858352ff66d46c39c0aea9033166276de0930935e02771d55682bb229","serverName":"ecocompute-estimator","capabilities":["tools"],"serverVersion":"1.1.0","protocolVersion":"2025-06-18"}},{"at":"2026-11-04T04:28:43.633Z","kind":"mcp_initialize","ok":true,"httpStatus":200,"latencyMs":54,"error":null,"detail":{"tools":[{"name":"estimate_quantization_energy","description":"Estimate the task-level energy change (ΔE% vs FP16) of weight-only quantization (NF4 or INT8) for a model of a given size on a GPU architecture class, from meas"},{"name":"get_quantization_crossover","description":"Get the model-size crossover for NF4/INT8 per architecture: below it quantization costs energy, above it saves. Crossovers belong to a card + software stack, no"},{"name":"list_data_coverage","description":"List what is actually measured (GPU architecture × precision × size range, NVML direct), what is borrowed/estimated, and where the gaps are. Use this before tru"},{"name":"search_measurements","description":"Search the measured anchors by model name, GPU name, architecture or size. Returns ΔE% vs FP16 per anchor with run counts — the raw evidence behind the curves."}],"toolCount":4,"toolsHash":"a5d211e858352ff66d46c39c0aea9033166276de0930935e02771d55682bb229","serverName":"ecocompute-estimator","capabilities":["tools"],"serverVersion":"1.1.0","protocolVersion":"2025-06-18"}},{"at":"2026-11-03T20:35:21.991Z","kind":"mcp_initialize","ok":true,"httpStatus":200,"latencyMs":304,"error":null,"detail":{"tools":[{"name":"estimate_quantization_energy","description":"Estimate the task-level energy change (ΔE% vs FP16) of weight-only quantization (NF4 or INT8) for a model of a given size on a GPU architecture class, from meas"},{"name":"get_quantization_crossover","description":"Get the model-size crossover for NF4/INT8 per architecture: below it quantization costs energy, above it saves. Crossovers belong to a card + software stack, no"},{"name":"list_data_coverage","description":"List what is actually measured (GPU architecture × precision × size range, NVML direct), what is borrowed/estimated, and where the gaps are. Use this before tru"},{"name":"search_measurements","description":"Search the measured anchors by model name, GPU name, architecture or size. Returns ΔE% vs FP16 per anchor with run counts — the raw evidence behind the curves."}],"toolCount":4,"toolsHash":"a5d211e858352ff66d46c39c0aea9033166276de0930935e02771d55682bb229","serverName":"ecocompute-estimator","capabilities":["tools"],"serverVersion":"1.1.0","protocolVersion":"2025-06-18"}},{"at":"2026-11-03T12:30:49.763Z","kind":"mcp_initialize","ok":true,"httpStatus":200,"latencyMs":112,"error":null,"detail":{"tools":[{"name":"estimate_quantization_energy","description":"Estimate the task-level energy change (ΔE% vs FP16) of weight-only quantization (NF4 or INT8) for a model of a given size on a GPU architecture class, from meas"},{"name":"get_quantization_crossover","description":"Get the model-size crossover for NF4/INT8 per architecture: below it quantization costs energy, above it saves. Crossovers belong to a card + software stack, no"},{"name":"list_data_coverage","description":"List what is actually measured (GPU architecture × precision × size range, NVML direct), what is borrowed/estimated, and where the gaps are. Use this before tru"},{"name":"search_measurements","description":"Search the measured anchors by model name, GPU name, architecture or size. Returns ΔE% vs FP16 per anchor with run counts — the raw evidence behind the curves."}],"toolCount":4,"toolsHash":"a5d211e858352ff66d46c39c0aea9033166276de0930935e02771d55682bb229","serverName":"ecocompute-estimator","capabilities":["tools"],"serverVersion":"1.1.0","protocolVersion":"2025-06-18"}},{"at":"2026-11-03T04:26:51.844Z","kind":"mcp_initialize","ok":true,"httpStatus":200,"latencyMs":372,"error":null,"detail":{"tools":[{"name":"estimate_quantization_energy","description":"Estimate the task-level energy change (ΔE% vs FP16) of weight-only quantization (NF4 or INT8) for a model of a given size on a GPU architecture class, from meas"},{"name":"get_quantization_crossover","description":"Get the model-size crossover for NF4/INT8 per architecture: below it quantization costs energy, above it saves. Crossovers belong to a card + software stack, no"},{"name":"list_data_coverage","description":"List what is actually measured (GPU architecture × precision × size range, NVML direct), what is borrowed/estimated, and where the gaps are. Use this before tru"},{"name":"search_measurements","description":"Search the measured anchors by model name, GPU name, architecture or size. Returns ΔE% vs FP16 per anchor with run counts — the raw evidence behind the curves."}],"toolCount":4,"toolsHash":"a5d211e858352ff66d46c39c0aea9033166276de0930935e02771d55682bb229","serverName":"ecocompute-estimator","capabilities":["tools"],"serverVersion":"1.1.0","protocolVersion":"2025-06-18"}}],"tools":[{"name":"estimate_quantization_energy","description":"Estimate the task-level energy change (ΔE% vs FP16) of weight-only quantization (NF4 or INT8) for a model of a given size on a GPU architecture class, from meas"},{"name":"get_quantization_crossover","description":"Get the model-size crossover for NF4/INT8 per architecture: below it quantization costs energy, above it saves. Crossovers belong to a card + software stack, no"},{"name":"list_data_coverage","description":"List what is actually measured (GPU architecture × precision × size range, NVML direct), what is borrowed/estimated, and where the gaps are. Use this before tru"},{"name":"search_measurements","description":"Search the measured anchors by model name, GPU name, architecture or size. Returns ΔE% vs FP16 per anchor with run counts — the raw evidence behind the curves."}],"package":null,"toolSurface":{"id":"ts_3b8fxrkyvx59","endpointId":"ep_hccx6aupxabc","hash":"a5d211e858352ff66d46c39c0aea9033166276de0930935e02771d55682bb229","toolCount":4,"serverName":"ecocompute-estimator","serverVersion":"1.1.0","protocolVersion":"2025-06-18","firstSeenAt":"2026-10-26T15:31:17.131Z","lastSeenAt":"2026-11-06T06:28:07.881Z","observations":34,"hiddenText":{"level":"none","total":0,"tools":0,"kinds":[],"findings":[]},"toolNames":["estimate_quantization_energy","get_quantization_crossover","list_data_coverage","search_measurements"],"distinctSurfaces":1,"surfacesWithHiddenText":0},"endpointFacts":[{"id":"ep_hccx6aupxabc","url":"https://ecocompute-estimator.zhanghongping1982.workers.dev/mcp","type":"mcp_streamable_http","factsCheckedAt":"2026-10-29T09:26:41.255Z","auth":{"observedAt":"2026-10-29T09:26:41.286Z","authRequired":false,"scheme":null,"resourceMetadata":null,"authorizationServer":null,"conformance":{"dpop":false,"rfc8414":false,"rfc9728":false,"pkceS256":false,"clientIdMetadataDocument":false,"dynamicClientRegistration":false}},"tls":{"observedAt":"2026-10-29T09:26:41.354Z","protocol":"TLSv1.3","chainValid":true,"chainError":null,"hostMatches":true,"subject":"zhanghongping1982.workers.dev","issuer":{"commonName":"WE1","organization":"Google Trust Services"},"validFrom":"2026-10-23T15:09:58.000Z","validTo":"2027-01-21T16:09:56.000Z","daysToExpiry":76,"sanCount":2,"fingerprint256":"CA:B8:E3:0F:FB:1D:52:2D:52:A0:EE:3E:EA:F2:AD:01:D9:56:DF:B8:B7:53:21:39:6C:EF:6F:D6:B7:D6:76:09"}}]},"verification":{"claimed":false,"claimedAt":null,"proofs":[]},"provenance":{"sources":[{"source":"mcp_registry","key":"io.github.hongping-zh/ecocompute-estimator","url":"https://registry.modelcontextprotocol.io/v0/servers/io.github.hongping-zh%2Fecocompute-estimator","firstSeenAt":"2026-10-26T14:19:04.640Z","fetchedAt":"2026-11-06T13:19:23.278Z","normalizedAt":"2026-11-06T13:19:23.278Z"}]},"firstSeenAt":"2026-10-26T14:19:04.640Z","updatedAt":"2026-11-06T06:32:36.899Z"}