Skip to content

Executable Science MCP server

Which MCP servers are Executable Science'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 2026-10-21

← Back to Executable Science

Whose server it is
  • Registry namescience.executable/executable-scienceverified in the MCP registry

The registry name sits under executable.science. The registry only lets a publisher use that namespace after proving control of the domain through DNS or HTTP.

When we asked itchecked 2026-10-21
  • Stateanswerschecked 2026-10-21
  • Tools15
  • Server calls itselfexecutable-science
  • Server version0.2.0
  • Protocol version2025-06-18

On 2026-10-21 the server answered initialize and tools/list without a sign-in. First answer in 380 ms.

Should an agent connect itNot proven either way yet
  • Whether its tools do the product's job has not been judged yet.

Once connected

  • Parameters are untyped or undescribed in 10 of 15 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.

Measured dailydaily checks: 7, the first on 2026-10-21
  • INITE MCP score91 of 100answers 100 · definitions 72 · context 100 · connects 100
  • Answered100% of days7 of 7 daily checks
  • Median first answer402 ms
  • Space in an agent's context≈ 1,777 tokensthe definitions of all its tools (15), which an agent carries on every turn
  • Tool definitionsC · 72 of 100
  • Typed results0 of 15 toolsdeclared outputSchema: results come typed, read once
  • Parameters without a description: in 10 of 15
  • Not saying whether it only reads or can destroy: 15 of 15

What an agent can do with it

  • not sure · Judge whether a small benchmark's success-rate interval is trustworthy
  • cannot · Find where two evaluation harnesses disagree on the same predictions
  • not sure · Trace a published claim back to its sources and a next check

Each of the product's use cases, put to Jev against this server's tool list: can an agent do it with these tools?

Worth a look

None of the five warning signs in these definitions.

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 →

Tools it offers15 tools · read 2026-10-21
  • get_public_preprint

    Read one exact, completed public preprint disclosure. A preprint is separate from journal acceptance; DOI registration, preservation and scientific verification are not inferred.

  • get_public_review_index

    Read the index and exact section counts of one frozen, published review disclosure. Review, DOI, preservation and publication have separate states.

  • get_public_review_item_bytes

    Read at most 32 KiB of one exact frozen public-review item, including a section or child item too large for a page. Follow next_action, match disclosure/version/section/parent/ordinal/item identity and digest across chunks, then verify…

  • get_resource

    Read one catalogued public resource by its identifier, including its recorded rights and provenance.

  • get_starter_source

    Download one original licensed starter ZIP as base64 with its SHA-256 and rights. Source has not run or been verified; submitting it still requires isolated ingestion.

  • get_starter_study

    Read one licensed starter study's frozen inputs, archive digest, environment, expected interpretation and unrun status.

  • get_task_template

    Read one frozen task template in full: its question, inputs, assumptions, comparison basis, stated limitations and revision digest. The digest pins the exact revision an attempt was prepared against.

  • list_public_review_coverage_page

    Read one bounded page of a published review's frozen literature-coverage declarations. A declaration is not independent evidence of every cited passage.

  • list_public_review_finding_events_page

    Read one bounded page of a published finding's frozen public events. Private or later responses are not included.

  • list_public_review_section_page

    Read one bounded section of a frozen published review disclosure. Finding events and literature coverage have separate child pages; follow next_action for complete history.

  • list_starter_studies

    List original downloadable source studies with explicit licenses, archive digests, pinned catalog environments and unrun evidence status.

  • list_task_templates

    List the bounded, frozen questions this platform publishes for attempt. Each carries a revision digest and its own stated limitations.

  • platform_status

    Report the current stage and, for every capability, whether it is available and the stated reason. Several capabilities are gated; this tool is the authoritative answer on which.

  • search_public_preprints

    Search only completed, currently accessible public preprint snapshots. Each result names an exact version and remains separate from journal articles, DOI registration, preservation and scientific verification.

  • search_resources

    Search public dataset, software, article, method and experiment listing metadata by title, summary, question, comparison, use case, source identity, rights and limits. Results retain source revision and readiness; they are not scientific…

How to connect

Add the server to your MCP client config. Claude Desktop, Cursor and most other clients read this shape:

{
  "mcpServers": {
    "executable-science": {
      "type": "http",
      "url": "https://executable.science/api/mcp"
    }
  }
}

Add it

Claude Code, in a terminal

claude mcp add --transport http executable https://executable.science/api/mcp

Add to CursorAdd to VS Code

claude.ai, ChatGPT and other clients: add https://executable.science/api/mcp as a custom connector in the settings.

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.