MCP HeartbeatTHE MCP OBSERVATORY
Public workspace

Explore

Evidence for your next connection.

Independent measurements of remote MCP servers. Every observation has a timestamp.

Read-only probes

MCP server · Other

AI Compute Radar

By aicomputeradar.dev Verified namespace

Which open models fit your GPU or Mac, measured. Model momentum, GPU rental prices, weekly pick.

Website
Connect to AI Compute Radar
Checks passed
12 / 22

2 failed · 8 not measured

Median response
170 ms

Median of successful probes over 7 days

30-day uptime
Not measured

Share of countable probes over 30 days, including daily rollups. Auth challenges are excluded.

Discovered tools
6

Tool definitions returned by this server

Independent monitoring

Health & checks

Status history

Responded

The server completed its last measured handshake. This is an observation, not a guarantee.

Last checked
Check summary by test group
Test areaPassedFailedUnmeasured
ConnectionCan a client connect and discover tools?
500
Reliability & speedDoes it respond consistently and quickly?
004
Access & authenticationWhat is required to use this server?
102
Tool qualityAre the tool definitions usable by an agent?
310
Publisher identityCan the publisher be traced to its source?
312

Unmeasured means there is not enough evidence to decide. How these checks work

Inspect individual checks 22 recorded

Core

  • Pass
    Reachable

    Latest probe returned HTTP 200 in 170 ms.

  • Pass
    Valid handshake

    protocolVersion 2025-11-25, serverInfo.name ai-compute-radar.

  • Pass
    Lists tools

    tools/list returned 6 tools.

  • Pass
    Current protocol

    Negotiated 2025-11-25. The bar is 2025-06-18 or newer.

  • Pass
    HTTPS

    The endpoint URL is https.

Performance

  • Not measured
    Fast (p50)

    Median latency 170 ms over 1 samples in 7 days. The bar is 1000 ms. Need at least 5 samples.

  • Not measured
    Consistent (p95)

    95th percentile 170 ms over 1 samples in 7 days. The bar is 3000 ms. Need at least 20 samples.

  • Not measured
    Reliable (uptime)

    1 of 1 attempts returned a tool list in 30 days (100%). The bar is 99%. Need at least 30 samples.

  • Not measured
    Stable tool list

    Fewer than two successful probes in 30 days, so tool-list stability is not measured.

Access

  • Pass
    Open access

    tools/list succeeded with no credential, and the record declares none.

  • Not measured
    Standard auth

    Auth is not required, so the OAuth metadata check does not apply.

  • Not measured
    Free tier cited

    No publisher quote has been checked.

Tools

  • Pass
    Tools described

    Every tool description is at least 20 characters.

  • Pass
    Valid input schemas

    Every inputSchema is a JSON Schema object.

  • Pass
    Parameters documented

    10 of 10 parameters have a description (100%). The bar is 80%.

  • Fail
    Safety hints

    6 tools set neither readOnlyHint nor destructiveHint.

Publisher

  • Pass
    Verified namespace

    Listed in the Official MCP Registry.

  • Pass
    Website matches publisher

    The website host matches the reverse-domain namespace.

  • Fail
    Source repo linked

    No repository URL is on file.

  • Not measured
    Repo owner matches

    This namespace is not an io.github login, so the owner comparison does not apply.

  • Pass
    Recently updated

    Updated 2026-10-27T17:17:24.765191Z, within 180 days.

  • Not measured
    Open-source licence

    No GitHub repository to ask about a licence.

Start using AI Compute Radar

Connect to this server

Choose your client, copy the configuration, then enable the server in your client’s MCP settings.

Run this command in your terminal, then restart Claude Code.

Claude Code

claude mcp add --transport http ai-compute-radar https://aicomputeradar.dev/api/mcp

Capabilities

Tools 6

find_fit

Which tracked models run on a given GPU or Mac: measured GGUF weights + computed context cache + runtime overhead versus usable memory. Returns the best recommendation and every verdict (EXCELLENT/GOOD/TIGHT/OFFLOAD_REQUIRED/NOT_RECOMMENDED/UNKNOWN) with plain-language reasons. Get hardware ids from list_hardware.

Discovered
gpu_prices

Median verified on-demand rental price per GPU class on Vast.ai (USD per hour), with min/p75 and offer counts, the collection timestamp, and per class the Rent Index: this week's median against last week and against the first week collected, a trend word, and the days excluded as marketplace glitches, plus RunPod's lowest posted on-demand price per class (a list price, not a median), Clore.ai's median for the same class (a second marketplace, never blended), and the AWS, Azure and Oracle Cloud pay-as-you-go list prices per GPU-hour, each with the instance type, VM size or bare-metal shape the price sits in (list prices, read four times a day, no statement about capacity). The index describes what prices did; it never forecasts.

Discovered
list_hardware

Curated GPU and Mac profiles the fit engine knows — ids, memory, usable memory after margins, bandwidth. Use an id with find_fit.

Discovered
price_history

Daily price history per GPU class and source, up to 90 days: the rows the Rent Index and the 30-day range are computed from. Marketplace days (Vast.ai, Clore.ai) are medians over the day's counted hourly medians with min, p75 and offer count; list-price days (RunPod, AWS, Azure, Oracle Cloud, Lambda, Nebius, CoreWeave, Hyperstack) carry the posted price. Each row says which kind it is; never blend the two. Describes what prices did; it never forecasts.

Discovered
trending_models

Tracked AI models ranked by Heat Score (0–100, weighted percentiles of measured Hugging Face/OpenRouter signals) with the raw signals, local-run facts (GGUF size, quantization) and links. Models still collecting a week of history have heat=null and rank after scored ones.

Discovered
weekly_pick

The current pick of the week: one tracked model chosen by a published rule (largest counted Heat Score rise among models that run comfortably on a consumer card of up to 24 GB), with the numbers frozen at selection time, a device-by-device fit ladder and the written report including its caveats. Pass week (e.g. 2026-w37) for a past issue. issue is null until the first issue is published.

Discovered

Recent measurements

Probe history

View status
Latest stored probes. Raw probe rows are kept for 14 days.
CheckedResultResponseTools
Connected170 ms6

Related servers

All alternatives

Same function and industry, ranked by checks passed then median latency.

  • Responded16 / 22 checks passed
    Catalyst evidence graph

    catalystproject.aiVerified

    Findings on what compounds and supplements do in the body, with evidence strength, quote and paper.

    OtherOther

    • Passed: Handshake
    • Passed: Tools listed
    • Passed: Updated recently

    Median latency 699 ms · Last checked

  • Responded15 / 22 checks passed
    wcway

    wcway.comVerified

    Find the nearest public toilet worldwide. Read-only tools, no account, no API key.

    OtherOther

    • Passed: Handshake
    • Passed: Tools listed
    • Passed: Updated recently

    Median latency 49 ms · Last checked

  • Responded15 / 22 checks passed
    Pick My Costume

    pickmycostume.comVerified

    Recommends Halloween costumes you can actually build, with step-by-step DIY build guides.

    OtherOther

    • Passed: Handshake
    • Passed: Tools listed
    • Passed: Updated recently

    Median latency 57 ms · Last checked

  • Responded15 / 22 checks passed
    Routine Run: write a routine

    routinerun.comVerified

    Write a step-by-step routine and get a link that opens it in Routine Run on iPhone. No sign-in.

    OtherOther

    • Passed: Handshake
    • Passed: Tools listed
    • Passed: Updated recently

    Median latency 62 ms · Last checked

Explore alternatives to AI Compute Radar