
EMT Compute MCP server
Which MCP servers are EMT Compute's own, from the official registry and from asking its own domain, whether they answer, and what they offer. A program checked each of these, nobody copied them from a product page.
read 2026-11-04
- Registry namecom.emtcompute/emt-computeverified in the MCP registry
The registry name sits under emtcompute.com. The registry only lets a publisher use that namespace after proving control of the domain through DNS or HTTP.
- Stateanswerschecked 2026-11-04
- Tools17
- Server calls itselfemt-compute
- Server version1.0.0
- Protocol version2025-06-18
On 2026-11-04 the server answered initialize and tools/list without a sign-in. First answer in 869 ms.
- Whether its tools do the product's job has not been judged yet.
- Checked on 1 of the 7 days it takes to judge reliability.
Once connected
- emt_cancel_experiment change data without a sign-in: anyone with the address can call them.
- Parameters are untyped or undescribed in 13 of 17 tools: check arguments before calling.
Said by rules from the measurements above. The likely first call is a model's pick from the tool list; read its description before relying on it.
- INITE MCP scoreafter a weeka score needs 7 daily checks: 1 so far
- Answeredmeasuring1 of the 7 daily checks it takes to print a share
- Median first answer380 ms
- Space in an agent's context≈ 2,935 tokensthe definitions of all its tools (17), which an agent carries on every turn
- Tool definitionsB · 87 of 100
- Typed results0 of 17 toolsdeclared outputSchema: results come typed, read once
- Parameters without a description: in 13 of 17
Worth a look
- Tools that change data, on a server that answers without a sign-in: emt_cancel_experiment
A program checks the server once a day: initialize and tools/list, no model. Tokens are the definitions' length at about four characters a token. The grade averages six checks of every tool's name, description, schema and annotations; it says how well the tools are described, not whether they work. The score weighs what was measured and nothing else: answering (25%), doing the product's job (25%), definitions (20%), context (10%), use (10%) and how readily a client connects (10%). A part that was not measured is left out, and a hidden instruction caps the score at 40. The state of vendors' MCP servers: method and numbers →
emt_catalogCurrent GPU offers. Prices in USD micro-units per GPU/hour; unknown metrics remain null. Paginated to keep responses small.
emt_clustersCurrent whole-node cluster offers; fabric declarations do not establish independent EMT qualification.
emt_templatesPublic GPU container templates for ordinary rentals.
emt_experiment_presetsList scientific presets or inspect one preset's parameters, equations, methods and custom-input schema. Does not run a calculation.
emt_ordersOwned GPU orders and lifecycle states. Requires read scope.
emt_experiment_jobsOwned experiment jobs, including reports and release state. Requires read scope.
emt_experiment_jobInspect an owned experiment and its authenticated result, output and calculation-report download links. Files are not sent into the model context.
emt_monitorMetrics and lifecycle for an owned order. Missing telemetry is unknown.
emt_quoteCreate a fresh expiring GPU quote without renting or charging.
emt_cluster_quoteQuote selected whole nodes on one fabric without renting.
emt_experiment_quoteQuote 1–4 nodes for an experiment; hours × hourly gives the GPU rental ceiling in micro-USD. Reserve is only an initial hold, not total cost.
emt_rentStart a PAID GPU rental. Requires explicit spending authorization, fresh quote, rent scope, credit and accepted terms. Store payload and key before requesting.
emt_rent_clusterStart a PAID whole-node cluster rental. Public SSH key only. Confirm cost and fabric limits with the user first.
emt_start_experimentStart a PAID scientific preset, customer script or command. Parameter sets distribute across quoted nodes; this does not automatically parallelize a single solver. Auto-continuation may rent additional segments up to its budget. Review…
emt_attach_cluster_experimentAttach work to an already rented dedicated 2–4-node cluster. Does NOT start SSH remotely: run emt_cluster.py locally with the returned job ID and your private key. Whole cluster stops on completion/cancel; billing continues during…
emt_cancel_experimentCancel an owned experiment and request release of all its GPU nodes or attached cluster. Unfinished work can be lost. Recheck releasing until false.
emt_terminateTerminate every node in an owned rental. Local disks are deleted; retained volumes and backups remain billed. Inspect status until release is confirmed.
How to connect
Add the server to your MCP client config. Claude Desktop, Cursor and most other clients read this shape:
{
"mcpServers": {
"emt-compute": {
"type": "http",
"url": "https://emtcompute.com/mcp"
}
}
}Add it
Claude Code, in a terminal
claude mcp add --transport http emtcompute https://emtcompute.com/mcpclaude.ai, ChatGPT and other clients: add https://emtcompute.com/mcp as a custom connector in the settings.
OpenAPI spec
85 paths, read 2026-11-04. Open the spec ↗
Each line is something a program found when it went looking: the registry entry, the server's own answer, a spec that parses. A missing line means we did not find it, not that it does not exist.
Before we count this visit
Analytics and ad measurement tell us which pages and which of our ads do their job: Google Analytics and Google Ads, the Meta pixel, Yandex Metrica, and one first-party id of our own. Nothing is stored until you answer, and declining changes nothing about the site.