# EasyNet — The Capability Network for AI Agents

Source: https://easynet.run/

Why EasyNet

## Your AI should know what is already there.

A useful script on your laptop. A model on the lab GPU. Project context in another session. Each is valuable, but a new agent only sees the small part connected to its current environment.

Recorded local function result

### Use the pricing rule. Don’t ask AI to recreate it.

When an assistant prepares a quote, it needs your pricing rules—not a guess. Publish the existing function so an authorized caller can send the plan and seat count, then use the returned amount.
[See the call and source](https://easynet.run/cases/validated-remote-pricing#recorded-quote-title)

#### Recorded request

Seats

48

Plan

team

Billing period

annual

#### Returned quote

Total

1,909,440 cents

Discount

336,960 cents

Sample pricing · recorded local CLI call, 4 Sep 2026. Not an AI-generated quote or a cross-device run. The record does not specify a currency.

People can find and use resources across devices and platforms.

## AI should be able to do the same.

EasyNet organizes connected capabilities around a common address, owner, and access policy. The resource keeps its own execution environment.

Select a resource, then switch agents to see how access changes.

### Your existing resources

### EasyNet Index
model.embed_text

A function beside a loaded model. Inputs and released results cross the call boundary.

Owner

research-team

Access

Allowed for this agent
[Read the connection guide](https://easynet.run/docs/connect/function)

### Your agents

Interactive illustration. Entries and access decisions are examples, not live network status.

### See Context and a function work together

An assistant queried a mapped folder, then sent the returned metadata to a function to prepare a handoff manifest.
[See the recorded workflow](https://easynet.run/docs/connect/context#recorded-context-function)

Synthetic same-device test through a restricted MCP bridge. One local Runtime identity; no file contents read. The fixture and function source were known to the assistant.

Start with a real task

## What could your next agent do?

A connected tool should return something useful. Explore source examples you can adapt to your own task.

### Prepare text for a retrieval workflow

A research agent needs vectors for document retrieval. Connect your existing embedding function; document indexing and search remain separate steps.

Input

A passage of text

Result

A vector

Public source exampleDeterministic 8-value stub. No model or GPU is used.
[See example and setup](https://easynet.run/cases/warm-gpu-model)

### Get selected fields, not the whole record

An agent needs a summary for a known record. Send its ID to the provider’s function and return only the fields that interface allows.

Input

A known record ID

Result

Short summary and allowed labels

Public source exampleTwo synthetic records, exact-name replacement and a 160-character limit. Not general anonymization.
[See example and setup](https://easynet.run/cases/data-resident-ai)

### Request a few images, not an endless stream

An agent needs a bounded visual sample. Request a frame count through one operation, which finishes after returning those frames.

Input

A frame count

Result

A finite sequence of images

Public source exampleGenerates 1–10 grayscale frames. No camera driver or device pairing.
[See example and setup](https://easynet.run/cases/paired-device-camera)

## What is your AI missing?

Start with one tool, one source of context, or one device. Find its connection path and see what you can set up today.
[Choose what to connect](https://easynet.run/docs/connect)[Talk through your scenario](https://easynet.run/pilot)

Runtime, MCP, and Context guides are available. Guided function integration with a second caller is part of Technical Alpha.