Customising your agents

Choose the model each agent runs on, and give them your instructions, knowledge and skills.

Providers and models

The Agents page lists every provider and the models it offers, and shows which are available right now. Each agent, and the assistant, runs on the platform default unless you pick another model; a provider that is unavailable falls back to the default automatically. The platform's own prompts and model map are never changed by your choices.

Standing instructions

Customize an agent to add standing instructions: text added to that agent's prompt on every call in your workspace. Use it for preferences the agent should always respect, such as a venue, a timeframe, a leverage ceiling or a style of hypothesis.

Knowledge files

Upload reference files on the Knowledge tab: notes, research, playbooks, data sheets, in text, Markdown, PDF, Word, EPUB or HTML. The text is extracted once and shown back to you, then attached to any agent with a switch. Attached files are inlined into the agent's prompt, up to fifty thousand characters per file and two hundred thousand per call.

Skills

A skill is a reusable procedure with a name, a one-line description of when it applies, and step-by-step instructions in Markdown. Write skills on the Skills tab or import a Markdown file, then attach them to agents. The agent applies a skill when the task matches its description.

Where it applies

Instructions, knowledge and skills apply to every pipeline call for your runs and to the assistant's replies to you. They refine the platform rules and never lift a safety, fee, venue, risk or output-format rule.

Operator status

Administrators also see each agent's live operator status on the same page: tasks completed, errors, uptime and last activity.