Find the answer in the work
Search across tasks, pages, projects, boards, forms, comments, sprints, and meetings. Agent Kit resolves ordinary names, so people do not need to know IDs or commands.
AI agents that can actually do the work
Ask in plain English. Superthread can find information, create and update work, and return real tasks, pages, boards, and projects—not a wall of text.
Create an agentIt does not stop at an answer
Agent Kit is the shared capability layer behind the Superthread agent, custom agents, our public CLI, and MCP. It gives AI more than 100 structured ways to work with the system where your team’s knowledge and execution already live.
Use it inside Superthread or connect tools such as Claude and Codex through the CLI and MCP server.
Take action, not just notes
Create and update tasks, pages, projects, boards, forms, comments, checklists, sprints, and agents. More than 100 structured actions are available through one capability layer.
Return real Superthread objects
Results come back as native task rows, page tiles, boards, sprints, comments, checklists, and search results that people can open and keep working with.
“Atlas is slipping by two weeks. Move the unfinished work, preserve the gaps between tasks, and show me what changed.”
An automation engine that understands the work
Give the agent a goal, then wake it up from the moments that matter. It gets the event and the surrounding Superthread context, not just a webhook payload.
Ask or mention
Start a chat when you need something, or @-mention an agent in a comment where the context already lives.
Run on a schedule
Run daily checks, weekly client updates, monthly reviews, or any recurring job in the right timezone.
React to discussion
Trigger from comments on tasks, projects, or pages, including comments within a board, sprint, or roadmap.
Start when work arrives
React to a form submission, a task entering a board or sprint, or a page or board being added to a space.
Watch the properties that matter
Run when status, assignees, list, priority, tags, sprint, project, dates, estimate, or other task properties change.
Scope every trigger
Listen everywhere, inside one space, board, sprint, or project, or only on one specific task or page.
Powerful does not mean uncontrolled
Give the agent a workspace role and limit it to the spaces where it should work. Then decide who can run it, improve it, manage it, or inspect other people’s chats.
- Can use
Run the agent without changing how it is configured.
- Can edit
Improve its instructions and settings as the job evolves.
- Full access
Manage the agent, its access, and the spaces where it can work.
- View all chats
Control separately who can inspect everyone’s runs, not only their own.
Make the agent part of the workflow
The best first agent has a clear job, useful context, and an obvious moment to run.
-
Describe the job
Write clear instructions in ordinary language. Tell the agent what a good result looks like and how you want it to behave.
-
Choose its context
Give it access to the spaces where the relevant projects, tasks, pages, and meeting notes live.
-
Choose when it runs
Use it on demand or connect it to schedules, comments, forms, new work, and property changes.
-
Review and improve
Inspect what it changed, leave feedback where the work lives, and refine the instructions as the job evolves.
Agents can help you build better agents
The system agent can create and configure other agents. That makes a deliberate feedback loop possible without removing the human who decides what “good” looks like.
-
An agent produces the work
It creates a task, page, report, plan, or another result inside Superthread.
-
A person judges the result
The reviewer comments on the actual output. The feedback stays attached to the work.
-
The system agent carries it forward
The Superthread agent can pick up that feedback and update the original agent’s instructions for the next run.
Built for agencies and internal service teams
Start where requests pile up, context is scattered, and good people spend too much time keeping the system tidy.
Digital agencies
Give every account team an operator that understands the client, the brief, the plan, the meetings, and the hours behind the work.
- Shift unfinished work when a client timeline moves
- Prepare weekly client updates from tasks, decisions, and meetings
- Check new briefs for missing detail before delivery starts
- Surface blockers, scope drift, and overdue decisions early
Internal service teams
Turn an intake portal into an intelligent front door for IT, People, Finance, Legal, Operations, and every team that handles requests.
- Answer common questions before a form is needed
- Find, fill, and submit the right request form
- Route work and chase missing context automatically
- Monitor queues and react to comments, statuses, and dates
Questions about AI agents
-
How are Superthread agents different from a chatbot?
They can do more than produce text. Agent Kit gives them more than 100 structured actions across tasks, pages, projects, boards, forms, comments, checklists, sprints, members, and agents. Their results can render as native Superthread objects that you can open and continue working with.
-
What is the difference between the Superthread agent and a custom agent?
The system-level Superthread agent is a general-purpose assistant for finding information and carrying out work across your workspace. Custom agents have a defined job, instructions, access, and triggers, so they can take responsibility for repeatable workflows.
-
How can an agent be triggered?
Start one from chat, @-mention it, run it on a schedule, or trigger it from comments, form submissions, new tasks, pages and boards, and detailed task-property changes such as status, assignee, priority, tags, project, dates, and estimate.
-
Can I control an agent’s access?
Yes. Choose its workspace role and the spaces where it can work. Give people Can use, Can edit, or Full access, and separately decide whether they may view everyone’s chats or only their own.
-
Can agents create or improve other agents?
Yes. The system-level Superthread agent can create and configure agents, including their instructions, access, tools, and triggers. You can build a deliberate feedback loop where one agent produces work, a person comments on it, and another agent uses that feedback to improve the original agent’s instructions.
-
Can agents be shared with a team?
Yes. Share an agent with specific people or make it available to the workspace. Access levels control who can run it, edit it, manage it, or inspect other people’s runs.
Get started in minutes,
import your tasks
Superthread is a drop-in replacement for your existing project management tool
