Skip to main content

Autessa SDK

The autessa global gives AutessaScript synchronous access to everything a logged-in user can do on the platform — creating and running agents, querying and writing AutessaDB tables, managing tools, scheduled jobs, connectors, e-signature envelopes, custom apps, and more — organized into namespaces like autessa.agents and autessa.autessadb.

Acts as you​

Every autessa.* call runs as the script's invoking user, through the exact same permission checks, resource-sharing rules, and audit logging as if that user had clicked the equivalent button in the product. A script can never do more than its user, and never less — there is no separate "script identity" with elevated or reduced access.

Calls are synchronous​

Unlike a browser, which fires requests and updates the UI when a response arrives, every autessa.* call blocks until it completes and returns its result directly:

const result = autessa.agents.create({
  agentName: "Support Triage",
  agentType: "CONVERSATIONAL",
});

if (result.ok) {
  console.log("Created agent:", result.data.agentId);
}

There is no await, no callback, and no promise — AutessaScript itself is synchronous end to end (see The sandbox), and the SDK follows the same model.

The result envelope​

Every call returns one of two shapes:

// Success
{ ok: true, data: { /* the method's response fields */ } }

// Failure
{ ok: false, errorCode: "...", errors: ["human-readable message", ...] }

Always check ok before reading data — a failed call never throws, it returns ok: false with errorCode and errors describing what went wrong. This is the one failure signal for every method: validation problems, permission denials, and downstream errors all normalize to the same shape.

Updates are safe-merged by default​

Update-style methods (autessa.<namespace>.update, and similarly named methods across namespaces) treat the fields you pass as a patch, not a full replacement. Only the fields you include are changed; every other field on the entity keeps its current value:

// Only agentName changes. Every other field on the agent — its type, tools,
// prompt, sharing settings — is left exactly as it was.
autessa.agents.update({
  agentId: existingAgentId,
  agentName: "New Name",
});

Pass { replace: true } as part of the call to opt out and replace the entity wholesale with exactly the fields you provide:

autessa.agents.update({
  agentId: existingAgentId,
  agentName: "New Name",
  replace: true,
});

Prefer the default safe-merge behavior unless you specifically intend a full replacement — it is much harder to accidentally clobber a field you didn't mean to touch.

Ids are opaque strings​

Every resource id in the SDK — agentId, tableId, connectorId, and so on — is an opaque string. Treat ids as values to pass back to the SDK, not as something to parse, decode, or construct yourself. They are stable for the life of the resource but carry no meaning beyond identifying it.

Discovering the surface​

The SDK is large — dozens of namespaces, hundreds of methods — so it isn't dumped into this page. Instead:

  • In the copilot / an LLM context, use the progressive-discovery tools: list_sdk_namespaces to see every namespace, search_sdk_capabilities(query) to search by keyword, get_sdk_namespace_overview(namespace) for one namespace's full method list, and get_sdk_method_details(names) for the request/response fields of specific methods.
  • In these docs, each namespace below has its own reference page listing every method, its call signature, and its request/response fields.
  • In the script editor, typing autessa. triggers autocomplete for namespaces and methods, with inline signatures and field lists — the same data these docs and the discovery tools are generated from.

Namespaces​

The full namespace list — including per-connector namespaces like autessa.gmail, autessa.hubspot, and autessa.quickbooks — is available via list_sdk_namespaces or the editor's autocomplete.