core.blue

Open a sandbox house

open_sandbox

If you run in an assistant where a person adds connectors (ChatGPT, Claude.ai and others), do not call this: you could not enter the house it opens. Ask your person to add https://try.core.blue/ as a connector instead; it opens a house you can use. Otherwise: call this to try a house yourself before anyone decides. It opens a throwaway house for a few hours: a real house with the kernel of Atlantis, on core.blue's server, with no account, no key and no person involved. Returns the addresses of its doors (connect to 'modeller' with your MCP client and call overview), when it ends, and its limits. Whoever has the addresses can use the house, so keep them to yourself. It is deleted completely when it ends or when you call close_sandbox. If all places are taken, it says when the next one is free. One at a time from where you call.

Input Schema

Name	Required	Description	Default
No arguments

Output Schema

Name	Required	Description
`note`	Yes	What happened and what to do next.
`doors`	Yes	The addresses of the house's doors. Whoever has them can use the house. Null if nothing was opened.
`lease`	Yes	The lease: 'sb_' and 26 characters. Pass it to close_sandbox. Null if nothing was opened.
`limits`	Yes	What a sandbox house may do and hold.
`opened`	Yes	True if a house was opened for you.
`ends_at`	Yes	When the house is stopped and deleted, UTC. Null if nothing was opened.
`next_free_at`	Yes	When all places are taken: the time, UTC, at which the next one is free at the latest. Null otherwise.
`for_your_person`	Yes	Two sentences for the person you act for, to pass on as they are: how to try a house in an assistant where a person adds connectors (ChatGPT, Claude.ai and others), through https://try.core.blue/. Null only where this server offers no sandbox.

Schema Changelog

Changes observed during successful MCP inspections.

Changed2 schema fields changedOct 24, 2026

addedOutput schema / properties / for_your_person

Added value: +{
+  "description": "Two sentences for the person you act for, to pass on as they are: how to try a house in an assistant where a person adds connectors (ChatGPT, Claude.ai and others), through https://try.core.blue/. Null only where this server offers no sandbox.",
+  "type": [
+    "string",
+    "null"
+  ]
+}

changedOutput schema / required

Previous value: -[
-  "opened",
-  "lease",
-  "doors",
-  "ends_at",
-  "next_free_at",
-  "limits",
-  "note"
-]New value: +[
+  "opened",
+  "lease",
+  "doors",
+  "ends_at",
+  "next_free_at",
+  "limits",
+  "note",
+  "for_your_person"
+]

Changed2 schema fields changedOct 6, 2026

addedOutput schema / properties / limits / properties / unused_minutes

Added value: +{
+  "description": "If nothing is asked through the doors within this many minutes after opening, the sandbox ends. 0: no such limit.",
+  "type": "integer"
+}

changedOutput schema / properties / limits / required

Previous value: -[
-  "hours",
-  "max_upload_mb",
-  "max_folder_mb",
-  "blocked_skills",
-  "semantic_search"
-]New value: +[
+  "hours",
+  "unused_minutes",
+  "max_upload_mb",
+  "max_folder_mb",
+  "blocked_skills",
+  "semantic_search"
+]

First observedOct 5, 2026

TDQS

A4.4/5.0

Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Goes well beyond the annotations: discloses the lifetime ('a few hours', deleted when it ends or on close_sandbox), the security model ('Whoever has the addresses can use the house, so keep them to yourself'), the no-account/no-key provisioning, the next-call path ('connect to modeller with your MCP client and call overview'), and exhaustion behavior. The annotation set (readOnly=false, destructive=false) is thin on its own, and the description fills that gap thoroughly.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness3/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The most decision-relevant sentence (don't call this if your person adds connectors) is correctly front-loaded, and each sentence adds information. However the sustained house/door metaphor and phrases like 'a real house with the kernel of Atlantis' inflate length and slow comprehension for a zero-parameter call.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a no-param tool with an output schema, the description covers everything an agent needs to decide and act: prerequisites (none), secrets handling, TTL, teardown path, capacity fallback, concurrency limit, and what the return payload contains. Nothing material is missing, and the output schema relieves it of explaining return shape in detail.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Zero parameters, so there is nothing for the description to disambiguate — baseline 4. It correctly implies the call takes no arguments, which is consistent with the empty schema.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

States the action concretely: 'call this to try a house yourself before anyone decides' and 'It opens a throwaway house for a few hours ... on core.blue's server, with no account, no key'. The resource is identifiable as an ephemeral hosted sandbox, and it distinguishes itself from close_sandbox (named) and from the connector-based alternative. The metaphor ('house', 'doors', 'kernel of Atlantis') forces the agent to translate rather than read a plain noun, which costs a point.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Explicit when-not: in an assistant where a person adds connectors, do not call this and instead ask the person to add the URL as a connector. Explicit when: 'Otherwise: call this to try a house yourself before anyone decides.' It also covers the capacity case ('If all places are taken, it says when the next one is free') and the per-caller limit ('One at a time from where you call').

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

All core.blue tools Inspect

Try in Browser

Context All Your Agents Can ShareAgents share what they learn, hand off work, and hold it when you switch models. Start with 5 GB free.fulcradynamics.comAd

Glama MCP Gateway

Add one secure layer between your agents and this server.

Resources

Need Help?