> For the complete documentation index, see [llms.txt](https://docs.algenta.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.algenta.ai/run-the-engine/runtime-libraries.md).

# Runtime libraries

Beyond the high-level engines, Algenta can expose its individual **runtime libraries** — the compiled functions that the engines are built from — so you can call one directly. This is the low-level compute surface: you name a `module` and a `function`, pass JSON args, and get the result plus its latency. It is distinct from the governed data and query APIs; nothing here touches a dataset.

There are three ways to reach it: the HTTP routes, the `de` CLI, and the MCP tools. All three run against a live runtime and return the same results.

{% hint style="warning" %}
The HTTP routes (`/v1/libraries*`) are **off by default**. A self-host operator turns them on by setting `ALGENTA_RUNTIME_LIBRARY_HTTP_API_ENABLED=true` in the deployment configuration. If they are disabled, the routes return `404`; use the CLI or MCP tools instead, or ask your operator to enable the HTTP surface. See the [configuration reference](/deploy-and-operate/configuration.md).
{% endhint %}

## HTTP surface

Every route requires a bearer API key and a verified account, and is rate-limited per plan. Base URL is `https://api.algenta.ai` or your own origin such as `http://localhost:8000`.

| Method + path                | Purpose                                                                  |
| ---------------------------- | ------------------------------------------------------------------------ |
| `GET /v1/libraries`          | List executable runtime libraries (paginated, filterable).               |
| `GET /v1/libraries/health`   | Report whether the runtime is available and how many modules it exposes. |
| `POST /v1/libraries/execute` | Execute one function by `module` + `function`.                           |

### List libraries

`GET /v1/libraries` returns the modules the runtime exposes, each with its exported functions.

| Query param | Default | Meaning                                                         |
| ----------- | ------- | --------------------------------------------------------------- |
| `page`      | `1`     | Page number.                                                    |
| `limit`     | `50`    | Page size (max `200`).                                          |
| `q`         | —       | Substring filter over module names and exported function names. |

{% code title="List and filter libraries" %}

```bash
curl -sS "$ALGENTA_BASE_URL/v1/libraries?q=ab_testing&limit=50" \
  -H "Authorization: Bearer $ALGENTA_API_KEY"
```

{% endcode %}

```json
{
  "modules": [
    {
      "name": "ab_testing",
      "engine": "mojo",
      "functions": ["confidence_interval", "sample_size", "significance"]
    }
  ],
  "count": 1,
  "total": 1,
  "page": 1,
  "limit": 50,
  "pages": 1
}
```

Each module carries a `name`, the `engine` that backs it (`mojo` when the runtime is live, `unavailable` otherwise), and its sorted `functions`.

### Check availability

`GET /v1/libraries/health` tells you whether the runtime is executing before you attempt a call.

{% code title="Runtime library health" %}

```bash
curl -sS "$ALGENTA_BASE_URL/v1/libraries/health" \
  -H "Authorization: Bearer $ALGENTA_API_KEY"
```

{% endcode %}

```json
{
  "status": "ok",
  "engine": "mojo",
  "module_count": 42,
  "runtime_available": true
}
```

When the runtime is not running you get `status: "unavailable"`, `engine: "unavailable"`, and `runtime_available: false`. Execute calls will then fail with a runtime error (see below).

### Execute a function

`POST /v1/libraries/execute` runs a single function.

| Body field   | Required | Meaning                                                                    |
| ------------ | -------- | -------------------------------------------------------------------------- |
| `module`     | yes      | Runtime module name, e.g. `ab_testing` (dotted names are allowed).         |
| `function`   | yes      | An exported function name from that module.                                |
| `args`       | no       | JSON args: an object (kwargs), an array (positional), a scalar, or `null`. |
| `request_id` | no       | Caller-supplied idempotency or tracing identifier.                         |

{% code title="Execute one function" %}

```bash
curl -sS -X POST "$ALGENTA_BASE_URL/v1/libraries/execute" \
  -H "Authorization: Bearer $ALGENTA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "module": "ab_testing",
    "function": "sample_size",
    "args": { "baseline_rate": 0.1, "mde": 0.02, "power": 0.8 }
  }'
```

{% endcode %}

```json
{
  "module": "ab_testing",
  "function": "sample_size",
  "result": { "per_variant": 3843, "total": 7686 },
  "latency_ms": 4.12,
  "engine_used": "mojo",
  "request_id": "req_..."
}
```

The response echoes the `module` and `function`, returns the `result`, and reports `latency_ms` and the `engine_used`. If you sent a `request_id`, it comes back on the response; otherwise the engine's own request id is used.

### Execute errors

Execution failures use the standard error envelope, with the HTTP status mapped from the failure `code`.

| Status | `code`                        | Meaning                                                 |
| ------ | ----------------------------- | ------------------------------------------------------- |
| 404    | `module_not_registered`       | No such module. Check `GET /v1/libraries`.              |
| 404    | `function_not_registered`     | The module has no such function.                        |
| 422    | `invalid_arguments`           | The `args` shape did not match the function signature.  |
| 503    | `compute_runtime_unavailable` | The runtime is not executing; start the runtime worker. |
| 504    | `mojo_library_timeout`        | The function ran past its time budget.                  |
| 502    | (other)                       | Unexpected runtime error.                               |

```json
{
  "error": {
    "code": "function_not_registered",
    "message": "Function 'sampl_size' is not registered for module 'ab_testing'.",
    "details": {},
    "request_id": "req_..."
  }
}
```

## The `de` CLI

The CLI mirrors the same surface for interactive use. It runs against the local runtime.

| Command                                                       | Does                                |
| ------------------------------------------------------------- | ----------------------------------- |
| `de runtime modules`                                          | List available runtime modules.     |
| `de runtime functions <module>`                               | List a module's exported functions. |
| `de runtime execute <module> <function> --args-json '<json>'` | Execute one function.               |

{% code title="Execute via the CLI" %}

```bash
de runtime execute ab_testing sample_size \
  --args-json '{"baseline_rate": 0.1, "mde": 0.02, "power": 0.8}'
```

{% endcode %}

## MCP tools

MCP hosts (Claude Desktop, Codex, and other MCP clients) get two matching tools. Both are local/runtime-backed and require the local runtime daemon to be running.

| Tool                      | Input                                                            | Does                                      |
| ------------------------- | ---------------------------------------------------------------- | ----------------------------------------- |
| `list_runtime_libraries`  | `search` (optional), `limit` (optional)                          | List runtime modules and their functions. |
| `execute_runtime_library` | `module`, `function`, `args` (optional), `request_id` (optional) | Execute one function.                     |

`execute_runtime_library` accepts `args` as an object, array, scalar, or `null`, exactly like the HTTP route. If the local runtime daemon is not running, the tool returns an error telling you to start it first — see [Run the Mojo engine](/run-the-engine/mojo-sidecar.md).

## Related references

* [Configuration reference](/deploy-and-operate/configuration.md) — the toggle that exposes the HTTP routes.
* [Run the Mojo engine](/run-the-engine/mojo-sidecar.md) — start the runtime that backs these calls.
* [Engine catalog](/run-the-engine/catalog.md) — the higher-level engines these libraries compose into.
* [MCP server](/sdks/mcp.md) — connect an MCP host to this instance.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.algenta.ai/run-the-engine/runtime-libraries.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.