Rykon AI Help
How can we help
Clear guides for every part of Rykon AI, from your first task to an agent you trust to act on its own.
Loading the articles
Popular:Run your first taskApprove an actionStop everythingMake an API key
Browse by topic
10 topics, 31 articles, each checked against the product as it is today.
Getting started
3 articles
Create an account, find your way round the console and run your first task.
Agents
4 articles
Give an agent a brief, a model and tools, start from a template, change or archive it.
Runs
5 articles
Start a task, watch it live, read the transcript and the result, and stop it.
Approvals and autonomy levels
4 articles
Risk classes, what each level may do alone, and how to approve, edit or reject an action.
Safety and sandboxes
3 articles
What a run can and cannot reach, kill switches, and the record nobody can edit.
Runner hosts
2 articles
Where your runs execute: shared sandbox nodes, a machine of your own, or the self-hosted runner on yours.
Usage and spend
2 articles
What a run costs, where to see it, and the daily ceiling that keeps spend bounded.
Team and roles
2 articles
Owners, admins, members and viewers, and what each can do.
API keys
4 articles
Make a key with only the scopes it needs, keep it safe, revoke it.
Account and security
2 articles
Signing in, your password and two-step sign-in, the audit log, your data and closing an account.
Popular articles
Where most people start.
- Getting startedWhat Rykon AI does, and where to find each partAI agents that do bounded work inside rules you set, with a person deciding anything above their level and a record of every action.
- Getting startedCreate your Rykon AI accountSign up with your work email, confirm it with the link we send, then name the account. You become its owner.
- Getting startedRun your first taskDescribe a task on Runs, start it, and watch it work, answered directly or in a sandbox of its own, until it finishes.
- AgentsCreate an agentAn agent is a name, a brief, a model, tools from the catalogue and an autonomy level. It starts at the most careful level.
- RunsStart a runA run is one task. Choose an agent or run a bare task, and set its turns and deadline. Simple tasks are answered directly; code, a shell or files get a sandbox.
- RunsTwo lanes, and why a run waitsSimple tasks are answered directly, with no sandbox. Work that needs code, a shell or files gets a sandbox of its own. A run that waits always says why.
- Approvals and autonomy levelsRisk classes and autonomy levelsEvery tool has a class from R0 to R4. An agent acts alone up to its level’s ceiling; above it, a person decides. R4 always waits.
- Approvals and autonomy levelsApprove, edit or reject an actionAn approval shows the exact action the agent wants to take. Approve it as it is, edit it first, or reject it with a note.
- Safety and sandboxesWhat a run can and cannot reachSimple answers come straight from the control plane, which never runs code. Anything that runs code, a shell or files gets a fresh, locked-down sandbox whose network reaches one place, the relay. Your keys never enter it.
- Safety and sandboxesKill switches: stop a run, an agent or everythingStop one run, pause one agent, or stop every agent on the account at once. Everything they cover is refused until released.
- Usage and spendWhat a run costs, and where to see itEvery model call is metered: tokens in, tokens out and cost. See them per run, per day and per agent under Usage.
- Team and rolesRoles and what each can doOwner, admin, member or viewer. Members work with agents and runs; viewers read; owners and admins also manage people, keys and autonomy.
- API keysMake an API keyA key lets your own code use Rykon AI with exactly the scopes you give it. It is shown once.
Still need a hand
A person reads every message and replies by email. Tell us what you did, what you expected and what you saw instead.
Write to us
A person reads every message and replies by email. Tell us what you did, what you expected and what you saw. support@rykon.ai
Email supportDeveloper docs
The API, authentication with scoped keys, errors and pagination, with examples in curl, TypeScript and Python.
Read the docsSecurity
How runs are sandboxed, how keys are kept, and how approvals and the record work, written from the code.
How we protect you