For Agents
PreviewDesigned for the agentic web.
Every Weasel product will ship as an MCP server, so the AI agents acting on behalf of their humans can sign up, transact, and put our products to work. Currently in preview. Join the waitlist for early access.
Why agent-native
In a few years, most software will be used by agents, not humans.
The hardest problem in agentic commerce is not the AI. It is the millions of products and services that were never designed to be operated by anything other than a human with a mouse. Sign-up flows assume eyes on a screen. Payment requires a card present. Account recovery needs a phone.
Weasel Den is building every product in our portfolio with that assumption inverted. Agent-readiness is not a roadmap item we hope to get to. It is the foundation we start from. The same code path that serves a human in a browser will serve an agent through MCP, with appropriate authentication, scoping, and audit.
When the agentic web arrives in earnest, every Weasel product will already be there.
How it works
The shape of an agent-native product.
A simplified view of how an AI agent uses a Weasel product on behalf of its human.
Human
Grants the agent permission to act on their behalf, with scoped capabilities and spending limits.
AI Agent
Connects to the Weasel Den MCP server using credentials the human authorized.
Weasel MCP
Authenticates, applies the permission scope, and routes calls to the correct product.
Product
Receipt Weasel, Professor Weasel, or Weasel Life executes the action and returns a structured result.
Capabilities
PreviewWhat agents will be able to do.
A preview of the tools each Weasel product will expose. Final names and signatures may change before general availability.
Receipt Weasel
submit_receipt
Submit a receipt or purchase confirmation for price monitoring.
list_tracked_purchases
List the purchases the agent is currently watching for the user.
get_refund_status
Check the status of an in-flight refund request.
cancel_tracking
Stop monitoring a specific purchase.
Professor Weasel
create_assignment
Set up a new assignment with AI-use guidelines and scope.
add_students
Provision student accounts for an assignment.
list_chat_logs
Retrieve student chat logs for a given assignment.
set_guidelines
Update the AI-use guidelines for an active assignment.
Weasel Life
schedule_task
Create a new task on the user’s schedule.
query_calendar
Read the user’s upcoming commitments.
create_reminder
Set a reminder linked to a person, place, or context.
capture_decision
Record a small daily decision for future recall.
Security model
Permission, not impersonation.
Every agent connection to Weasel Den is bound to a specific human, with a permission scope that the human explicitly grants. Agents do not impersonate users. They act with delegated, time-bound, revocable authority.
Each product exposes its own capability scopes. A user can authorize an agent to track new purchases on Receipt Weasel without ever granting access to read past refunds. Granular by design.
Every call an agent makes is logged. Users can audit, revoke, and replay agent actions from a single console.
OAuth 2.1 with PKCE
Standard authorization code flow, designed for agent clients.
Scoped capability tokens
Each token carries only the permissions the human granted.
Spend ceilings
Per-token, per-product, and per-day limits on financial actions.
Auditable by default
Every agent action is logged and surfaced in the user’s console.
Get connected
Example configuration.
These configs are previews. Endpoints and authentication flows will be finalized before general availability.
{ "mcpServers": { "weaselden": { "url": "https://mcp.weaselden.ai", "auth": { "type": "oauth2", "client_id": "your_client_id" } } }}// Preview. SDKs and endpoints subject to change.
Early access
Build with Weasel agents first.
We are working with a small group of early agent developers before general availability. Drop us a line and tell us what you are building.
Common questions
Things people ask.
When does the MCP server actually ship?
Receipt Weasel is the first product targeted for MCP availability. We are not announcing public dates yet. Waitlist members will be notified before general availability.
What is MCP?
Model Context Protocol is an open standard for connecting AI agents to tools and data sources. It is supported by Claude, and increasingly by other agent frameworks. Weasel Den’s endpoints will follow the public MCP spec.
How does payment work when an agent acts for me?
Each user authorizes a spend ceiling per agent and per product. Agents cannot exceed those ceilings. Every paid action is logged and reversible within the product’s normal refund window.
What stops a malicious agent from abusing a Weasel product?
Every agent connection is bound to a real human account with explicit, scoped, revocable permission. Rate limits and anomaly detection apply at the MCP layer. Suspicious patterns terminate the session.
Will there be SDKs?
We will publish typed SDKs for Python and TypeScript at general availability. The raw MCP interface is available to anyone who wants to integrate without a SDK.
Is this open source?
The Weasel Den MCP servers will remain proprietary. Our public-facing documentation, example clients, and llms.txt feed are open and free to reference.