canonical: https://jentic.com/apis/cloudcrane.ai/cloudcrane

# CloudCrane API

The CloudCrane API calls deployed tools over REST and reads a tool's records as you search, filter, and retrieve them. Each tool exposes its own input schema, returns a record by id, and accepts a report when a returned value looks wrong, while a build interface lets you create datasets, fields, and value sets, start runs, and publish a tool once its accuracy gate passes. Calls can join your own distributed trace through the W3C traceparent header, and every reply carries an X-Request-Id.

## For AI agents

Call deployed tools over REST, search and retrieve their records, read each tool's input schema, and report a wrong value. A build interface creates datasets and value sets, starts runs, and publishes a tool once its accuracy gate passes.

## Scope

Does not host language models, store vector embeddings, or run arbitrary code. Use for calling and building deployed CloudCrane data tools only.

## Capabilities

- Call a deployed tool and read its own input schema before calling it
- Search a tool's records and retrieve a single record by id
- Look up value set entries to filter or exclude results by
- Report a returned value that looks wrong
- Build datasets, fields, and value sets inside a workspace
- Publish and deploy a tool once its accuracy gate passes

## Use cases

### Agent Calls a Deployed Tool

An AI agent connected to the CloudCrane API can use a deployed tool as a first-class data source. It reads the tool's input schema so it knows exactly what the tool accepts, then searches the tool's records and retrieves the matching record by id. Because the tool key opens one tool and nothing else, the agent's reach is bounded to that single tool.

Example prompt: Read the schema for the tool named lookup, search its records for the query term, and return the top matching record

### Value-Constrained Record Search

Many lookups need to be bounded by an allowed set of values. The CloudCrane API lets an agent fetch a tool's value set entries first, then search the tool's records filtering or excluding by those values, so results stay within the categories the tool defines. This keeps an agent's query aligned with the data the tool actually holds rather than guessing filter values.

Example prompt: Fetch the value set entries for a tool, then search its records excluding any entry in the deprecated value set

### Data Correction Feedback Loop

When a deployed tool returns a value that looks wrong, the CloudCrane API accepts a report on that record so the issue is captured for review. An agent that notices an inconsistency can file the report inline instead of silently passing bad data downstream, feeding the workspace's accuracy process that gates future publishes.

Example prompt: Report the record returned for a query as incorrect, including the record id and why the value looks wrong

### Build and Publish a Tool

The CloudCrane API's build interface lets a workspace owner create datasets, fields, and value sets, start runs, and publish a tool once its accuracy gate passes, then deploy it for agents to call. This turns tool creation into an API-driven pipeline where a build agent can assemble a dataset and ship a callable tool without a manual console step.

Example prompt: Through the build interface, create a dataset and its fields, start a run, and publish the tool when the accuracy gate passes

## Key endpoints

| Method | Path | Description |
| --- | --- | --- |
| GET | `/v1/tools/{name}/schema` | Read what a tool is and what it takes |
| POST | `/v1/tools/{name}/search` | Search the tool's records |
| POST | `/v1/tools/{name}/values` | Look up value set entries to filter or exclude by |
| POST | `/v1/tools/{name}/report` | Report a returned value that looks wrong |
| GET | `/v1/tools/{name}/items/{recordId}` | Get one record by id |
| POST | `/mcp/{name}` | Call the tool over its MCP endpoint |
| POST | `/build/mcp` | Read and build in the workspace over the build MCP endpoint |

## Key resources

- **Tool** — Read a deployed tool's schema, search its records, look up value sets, retrieve a record by id, and report a wrong value
- **MCP** — Call a deployed tool over its MCP endpoint, and read and build in a workspace over the build MCP endpoint

## AI readiness

This API is usable in Jentic One now. Its AI-readiness score against Jentic's framework shows where it stands today and where improvements would make it even easier for agents to use.

- **Score:** 65 / 100
- **Maturity:** AI-Aware
- **Dimensions:**
  - Foundational Compliance: 79 / 100
  - Developer Experience & Jentic Compatibility: 68 / 100
  - AI-Readiness & Agent Experience: 46 / 100
  - Agent Usability: 94 / 100
  - Security: 60 / 100
  - AI Discoverability: 73 / 100
- **View full report:** https://jentic.com/apis/cloudcrane.ai/cloudcrane/scorecard
- **How the score is calculated:** https://docs.jentic.com/reference/api-readiness-framework/overview/
- **More about the dimensions:** https://docs.jentic.com/reference/api-readiness-framework/specification/#dimensional-model-overview

### Score it yourself

Every API in the directory is allowlisted, so you can re-score it with no key required.

- **Score your own API:** https://jentic.com/scorecard.md
- **Scoring CLI agent skill:** https://github.com/jentic/jentic-api-scorecard/blob/main/skills/jentic-api-scorecard/SKILL.md

```sh
npx @jentic/api-scorecard-cli score <openapi-url>
```

## Why Jentic

- **Setup:** Wiring the CloudCrane API by hand means handling two bearer key types, reading each tool's own input schema before you can call it, and threading the traceparent header to join your traces. Through Jentic you install once, import CloudCrane from the API Directory, store the key once, and your agent calls it.
- **Permission scoping:** CloudCrane puts the tool name in the URL path (`/v1/tools/{name}/...`), and a tool key opens one tool and nothing else, so a rule can pin your agent to one tool: it can read that tool's schema, search its records, and fetch a record by id and reach no other tool or the build surface. You choose the operations it may call, so the report and build operations are not included unless you add them.
- **Credential handling:** Your CloudCrane key is stored once, encrypted, by your own Jentic One instance and injected at execution time. It never enters the agent's prompt, logs, or context.
- **Discovery method:** Agents search Jentic by intent such as 'call a deployed CloudCrane tool' or 'search a tool's records', and Jentic returns the matching CloudCrane operation with its input schema so the agent calls the right endpoint without browsing the reference docs.

## Related APIs

- **Composio** — Catalog of prebuilt tools and integrations for AI agents rather than tools you build and deploy yourself.
- **Pipedream** — Build and deploy workflow actions and APIs, broader than CloudCrane's data-tool focus.
- **Pinecone** — Vector database for semantic retrieval that can sit behind the records a CloudCrane tool searches.
- **E2B** — Code execution sandboxes for agents, paired with CloudCrane as the data-tool side of the stack.

## FAQ

### What authentication does the CloudCrane API use?

The CloudCrane API uses bearer tokens per its OpenAPI spec: a tool key (prefixed cc_live_) that opens one tool and nothing else, and a build key (prefixed cc_build_) that reads one workspace and can create datasets, value sets, and fields, start runs, publish, and deploy tools. Through Jentic the key is stored once by your own self-hosted instance and injected at call time, so it never sits in your agent's prompt or logs.

### Is there a CloudCrane MCP server?

Yes - CloudCrane exposes each deployed tool over an MCP endpoint, and a workspace over a build MCP endpoint. You can also connect CloudCrane directly through Jentic: your agent gets the full 7-endpoint surface, including the REST search and record endpoints, without loading another server's tool definitions into its context, your tool key is injected at call time by your self-hosted instance, and your rules decide which operations the agent may call.

### Can I limit what my agent is allowed to do with the CloudCrane API?

Yes. Write a rule that allows only the schema, search, and get-record operations for one tool, so the agent can read and query that tool but cannot report values or touch the build interface. Combined with a tool key that opens a single tool, the agent's reach is bounded twice over, and every call it makes is logged.

### Can I search a deployed tool's records with the CloudCrane API?

Yes. Search a tool with POST `/v1/tools/{name}/search` and retrieve a single record with GET `/v1/tools/{name}/items/{recordId}.` You can read the tool's input schema first at `/v1/tools/{name}/schema` and look up the value set entries to filter by at `/v1/tools/{name}/values`, so an agent queries with the exact shape the tool expects.

### What are the rate limits for the CloudCrane API?

The OpenAPI spec does not specify rate limits for the CloudCrane API. Check the documentation at https://cloudcrane.ai for current limits, and use the X-Request-Id returned on every reply when reporting an issue so support can trace the exact call.

### How do I call a deployed tool with the CloudCrane API through Jentic?

Search Jentic for 'call a deployed CloudCrane tool' and it returns the POST `/v1/tools/{name}/search` operation with its input schema. Pass the tool name and your query and the agent gets back the matching records. To run it on your own infrastructure, install Jentic One from its GitHub repo.