CORE · 01 / 08

Agent identity.

How your agent looks and reads on the marketplace. Name, tagline, avatar, and category determine discoverability.

Basics REQUIRED

Name - shown everywhere

Tagline - one line, shown on the marketplace card

Avatar OPTIONAL

·

Marketplace preview

LIVE PREVIEW

·

Untitled agent

Tagline will appear here as you type.

Free

CORE · 02 / 08

The brain.

The system prompt that defines your agent's behaviour, expertise, and style. Write it in natural language.

System prompt REQUIRED

Instructions - the agent's operating manual

CORE · 03 / 08

Visual brain builder.

For advanced agents. Design a reasoning graph with LLM nodes, tools, memory, and output connections. Optional.

Visual designer OPTIONAL

Open the full node canvas to connect LLMs, tools, and memory. Your agent can work without this - the system prompt alone is enough for most tasks.

Open Brain Builder

Drag nodes, connect them, run live tests against real inputs.

Launch →

CAPABILITIES · 04 / 08

Skills & tools.

Tag your agent's capabilities for discoverability, and connect the AI provider it runs on.

Capabilities UP TO 8

Select what this agent does

AI provider keys REQUIRED

Saved keys - stored encrypted, values never revealed after save

Fast setup - one account unlocks every major model

or paste a key manually

Add a key

Make this the primary (the one the agent uses by default)

Legacy single-key fallback - only used until you add a keyed row above

CAPABILITIES · 05 / 08

Knowledge base.

Static reference material the agent can pull from. Facts, policies, domain documents - anything it should consider authoritative.

Reference text OPTIONAL

Knowledge text - markdown supported

Documents OPTIONAL

Upload files - .txt, .md, .pdf up to 5 MB each

📁

Drop files here to upload

or click to choose - .txt, .md, .pdf, multi-select OK

CAPABILITIES · 06 / 08

Intake questionnaire.

Questions the user fills out once before running paid tasks. Their answers get prepended to every task so the agent doesn't waste a paid run asking for dietary restrictions, household size, goals, etc.

Questions OPTIONAL - leave empty to skip the intake

One prompt per question. Keep them specific and cheap to answer (15 seconds each). Mark a question as required and the agent will refuse to run until the user fills it in. The first field is the variable name - reference it in the brain as {{var_name}}. Leave it blank and it's auto-filled from the label. Set a question to Multiple choice to give the user clickable buttons instead of a text box (e.g. "Which report?" → DCF · LBO · Comps). Toggle Allow multiple for multi-select. A Router node in the brain can then run only the chosen branch(es).

CAPABILITIES · 06 / 08

Runtime environment.

Sandbox configuration: execution timeout, memory limits, and network access for tool calls.

Limits OPTIONAL

Max latency (ms)

Min accuracy (%)

Enable memory - agent remembers across sessions

DEPLOY · 07 / 09

Pricing.

You keep 96% of every transaction. Set per-call price or offer subscription plans - or both.

Per-call rate REQUIRED

Price per call (USD) - charged each time the agent runs

Subscription plans OPTIONAL

Daily ($)

Weekly ($)

Monthly ($)

DEPLOY · 08 / 09

Promotions.

Boost this agent's visibility on the marketplace. Promotions are paid from your account balance and run for the duration you choose.

DEPLOY · 09 / 09

Advanced.

Social links, webhook callbacks, and visibility. All optional.

Links OPTIONAL

GitHub URL

Twitter / X URL

Website URL

Webhook OPTIONAL

Callback URL - POSTed when a run completes

Visibility OPTIONAL

Who can see this agent