Dynamic Functions

Dynamic Functions are your own serverless JavaScript, running securely inside your Portablemind workspace. They let you extend the platform without deploying any infrastructure: pull data from external APIs, transform records, run scheduled jobs, or react to things happening in your workspace — all from a code editor in the app.

Each function executes in a sandboxed runtime that is isolated to your workspace, with a configurable timeout, so experiments can't run away from you.

You'll find them in AI Studio, under Advanced / Dev → Dynamic Functions. Create one with New Function, or bring one in with Import Dynamic Function (functions export and import as JSON, which makes them easy to share between workspaces).

Writing a function

A function is plain JavaScript. Two things are in scope when it runs:

• params — the input parameters passed by whoever invoked the function.
• context — runtime context, including context.config (the function's own configuration object) and context.oauth.accessToken (an access token, when a shared OAuth configuration is connected — see below).

The editor starts you with a working template, and the value your code produces is returned to the caller. Each function has a Timeout (seconds) setting that caps how long a run may take.

Running functions

There are three ways a dynamic function runs:

• Manually — the Execute (manual run) action on the function's row runs it on demand with parameters you supply. Great while developing.
• On a schedule — configure recurring runs directly on the function: every N minutes, every N hours, daily at a set time, monthly on a day of the month, or a raw CRON expression for anything more exotic. Scheduled runs execute as the user you pick in Run As User, and the Active toggle turns the schedule on and off.
• On platform events (callbacks) — attach the function to a Callback Model and Callback Event so it fires automatically when that kind of record changes in your workspace, turning functions into lightweight automation hooks.

Every run is recorded — the Execution history action shows past runs with their results, so you can see what a scheduled or callback-driven function has been doing.

Connecting to external services (OAuth)

Functions that call external APIs can use a Shared Configuration (OAuth) — a workspace-level OAuth connection to an outside service. Once connected, your code receives a ready-to-use context.oauth.accessToken, so you never handle client secrets or token refresh yourself. This is the pattern for integrating accounting systems, CRMs, and other third-party APIs.

Dynamic functions and AI

Dynamic functions plug into the rest of the platform:

• AI agents can list and execute your dynamic functions as tools, which means you can hand an agent a custom capability just by writing one.
• Combined with schedules and callbacks, a function can feed fresh external data into the workspace that your agents and AI memory then work from.

Tip: If a capability is mostly instructions for an agent, package it as a skill; if it's mostly code that needs to run reliably (API calls, data transforms, scheduled jobs), make it a dynamic function. The two compose well.

For the underlying API — managing and executing functions programmatically — see the API guide.