Agent Operator Library
Structured prompts and workflows for repeatable AI-assisted tasks. Use the patterns as-is or adapt them to your own operating rhythm.
$39View offer ↗
CROSSINGKEY AGENT DOCK
Dock the task. Build the agent. Launch the work.
A professional station for human and machine agents: buy a starting capability, connect your own API, set the boundary, and move a reviewed result into the work that matters.
01 / THE DOCK
One station for the human decision and the machine action.
Dock a real task, define the operating context, and choose the right capability. The product brings the operating logic; you bring the resource and the judgment.
01
Dock the brief
State the task, source material, desired result, and what must remain human.
02
Build the boundary
Connect the API, choose the model, and set approvals before a run.
03
Launch the handoff
Review the trace, keep the result, and move it into your real workflow.
02 / DOCK MODULES
Ready-to-use packages for the dock. Custom agent work when the task needs your systems, data, or rules.
Local AI Field Guide
A practical guide to private, offline AI workflows for creators and small teams that want more control over their context.
$29View offer ↗
Business Systems Vault
Editable planning and operating templates that give an agent better inputs and give the reviewer a clearer frame.
$59View offer ↗
CrossingKey MCP
Machine-readable discovery, evaluation, payment, receipts, and bounded capabilities for agents that need a real service behind the interface.
From $0.01Use the MCP ↗
An agent for your operation.
Bring the task, tools, and approval boundary. We scope the smallest useful system and define what “done” means before implementation.
ScopedStart a brief ↗
02A / CAPABILITY DOCK
Agent capabilities that can be inspected, priced, tested, and called through the CrossingKey machine interface.
Capability registry
Connect the Agent Dock backend to load live capability metadata, pricing, permissions, and MCP tool bindings.
03 / BUILD STATION
Connect your API and build a controlled session.
This buyer workbench supports OpenAI, Anthropic, and OpenAI-compatible endpoints. Your key stays in memory for this session and is sent only to the provider you choose; it is not stored by CrossingKey.
BYOK / SESSION ONLY
WORKSPACE ACCESSPreview mode — connect Supabase to enable protected workspaces.
SESSION OUTPUTIDLE
Configure a provider, choose an agent template, and run a session. The output will appear here with a lightweight trace.
TRACEwaiting for a run
04 / LAUNCH CONTROLS
Launch with a visible boundary.
The station is designed around buyer-owned context and explicit control. It can help you think, transform, review, and prepare; it does not silently publish, spend, or change an external system.
READ-ONLY DEFAULT
Runs produce an answer and trace before any action is considered.
BUYER-OWNED ACCESS
You connect the provider and choose the model. CrossingKey does not ask for your account password.
HUMAN REVIEW
Outputs are drafts until you approve, edit, or deliver them yourself.
RECEIPT MINDSET
Each useful run should leave enough context for someone else to understand the result.
05 / LAUNCH CENTER
Pay for the module. Bring the resource. Launch the result.
Choose a package from the sales collection, connect the API you already control, and use the agent inside your own workflow. Custom work starts with a scoped brief and a clear launch boundary.
06 / DOCK INTAKE
Have a process worth turning into a launchable agent?
Send the task, the systems involved, and the boundary that must stay human.