Assist is a platform for deploying governed AI teammates that can work across company systems with role-based permissions, secure connectors, model choice, audit trails, and cost visibility.

What are sandcastles?

Sandcastles are purpose-built internal applications created on Assist. They let teams build custom workflows, dashboards, forms, and AI teammate interfaces without buying another SaaS tool for every problem.

How much does Assist cost?

Assist has a self-serve plan for teams that want to build their own AI teammates, a managed plan where Assist builds and manages three AI teammates for $5,000 per month, and custom Enterprise pricing for advanced security and deployment needs.

Do we need engineers to use Assist?

No. Assist is designed so operators and subject-matter experts can work with AI teammates safely. Engineers can extend the platform, but teams can start with managed AI teammates or self-serve workflows.

Assist docs

Documentation

Reference material for the Assist runtime: tools, AI teammates, skills, sandcastle apps, MCP servers, scheduled triggers, filesystems, permissions, and shared chats.

Start here

Follow the runtime path.

Most teams start with the tool layer, connect it to an AI teammate, then lock down permissions before expanding access.

StartUnderstand the tool modelTools are the reusable execution layer behind AI teammates, sandcastle apps, MCP servers, and scheduled jobs.AI teammatesConfigure a teammateLearn how prompts, tools, schedules, channel routing, and access controls fit together.GovernanceSet the permission boundaryReview users, groups, resource permissions, SSO, and the simulator before wider rollout.

Sections

All documentation.

58 reference pages and 27 playbooks across the core Assist product surface.

Tool Creation

Write TypeScript tools that connect Assist to external systems and shared workflow capabilities.

7 references5 playbooks

Tool Creation Use Cases and Playbooks Creating a Tool Testing a Tool Editing and Versioning +2 more Playbooks

Subagents

Configure AI teammates with focused prompts, tools, schedules, channel routing, and workspace controls.

9 references5 playbooks

Subagents Use Cases and Playbooks Creating a Subagent Adding Tools to a Subagent Delegating to a Subagent +4 more Playbooks

Agent Skills

Package repeatable AI workflows as governed skills that teams can import, approve, attach, and version.

8 references3 playbooks

Agent Skills Use Cases and Playbooks Importing a Team AI Workflow Skill Creating a Managed AI Workflow Skill Authoring and Approving Agent Skills +3 more Playbooks

Sandcastles

Build internal web apps through conversation, then manage data, permissions, and iteration workflows.

6 references4 playbooks

Sandcastles Use Cases and Playbooks Creating a Sandcastle Editing a Sandcastle Managing Sandcastle Data +1 more Playbooks

MCP Servers

Expose workspace tools to external AI clients over OAuth-authorized MCP connections.

6 references6 playbooks

MCP Servers Use Cases and Playbooks Creating an MCP Server Authorizing an MCP Client Managing MCP Servers +1 more Playbooks

Scheduled Triggers

Run AI teammates on a schedule and review each job run from the Tasks page.

4 references1 playbooks

Scheduled Triggers Creating a Scheduled Trigger Viewing Job Runs Troubleshooting Playbooks

Agent Filesystem

Give AI teammates a private persistent filesystem for notes, drafts, research, and reusable context.

6 references3 playbooks

Agent Filesystem Use Cases and Playbooks Enabling an Agent's Filesystem Disabling an Agent's Filesystem Sharing an Agent's Filesystem +1 more Playbooks

Workspace Permissions

Manage users, groups, ownership, resource access, SSO, and permission simulation.

7 references

Workspace Permissions Managing Users Managing Groups Managing Resource Permissions Using the Permissions Simulator +2 more

Chat Sharing

Share AI chat conversations with teammates for review, source checking, and reuse.

5 references

Chat Sharing Sharing a Chat Viewing a Shared Chat Managing Permissions Troubleshooting

Machine-readable docs are available at/docs.md