Skip to content

Elixir / OTP / terminal / local-first

Local agent.
Writes code.

Beamcore is a real terminal TUI for coding sessions. It keeps context, renders EEVA execution, writes focused functions, runs them locally, and returns changes through one timeline.

test - beam.smp -- -root ~/.beamcore/app -bindir ~/.beamcore/app/erts-16.4/bin -prog...
        }
      </style>
    </head>
    <body>
      <div class="container">
        <h1>404</h1>
        <p>Oops! The page you're looking for doesn't exist.</p>
        <p><a href="/">Return to Home</a></p>
      </div>
    </body>
  </html>
""")
EEVA
```elixir
File.write!("gallery.html", """
<!DOCTYPE html>
<html lang="en">
<head>
  <meta charset="UTF-8">
  <meta name="viewport" content="width=device-width, initial-scale=1.0">
  <title>Gallery</title>
  <style>
    body {
      font-family: 'Arial', sans-serif;
      background-color: #f4f4f9;
      
Ask BeamCore, describe a change
◢▣◣ F1 Agent F2 Chat mistral/mistral-medium-3-5 tok 3.3k/5.2k

Product

The terminal is the product.

No fake dashboard. Beamcore shows code, EEVA execution, file context, status, and results in the same screen.

01

Strong session

One durable loop carries task intent, files, output, and decisions across the run.

02

Generated functions

The agent can write small code paths to inspect, transform, validate, and report.

03

EEVA rendered

Execution steps appear as readable code blocks, not hidden background magic.

04

Local-first

It runs where the repo lives. Your machine, your data, your responsibility.

Runtime

User input. Nodes. Function. Result.

The session receives the task, prepares context nodes, generates a small focused function, runs it locally, and shows the outcome.

User input Node A Node B Generated function Result

Execution path

User input flows through context nodes into a generated function, local execution, and result.

User input flows to node A and node B, then into a generated function, local execution, and result

Why people will point at it

You decide the limits.

Use it, inspect it, constrain it, fork it, or refuse it. Just do not pretend it is a toy.

01

It writes code to work with code.

Beamcore can generate small functions for the task instead of only choosing from a fixed tool menu.

02

It shows the execution.

EEVA blocks, code, terminal status, and results are visible in the same product surface.

03

It runs where the repo lives.

Local-first means the real project is the input, not a sanitized product demo.

04

It ships without a guarantee.

Powerful software can do excellent work and unexpected work. That is the point of reading the source.

No warranty

Run it with your eyes open.

Beamcore can read, write, and execute code on your machine. It can produce excellent work and unexpected work. You are responsible for how you run it.

Read the source

Get started

Build the agent from source.

Beamcore is open source. Clone the agent, prepare dependencies, and launch the full-screen TUI.

Install from source
$ git clone https://github.com/beamcore/agent.git
$ cd agent
$ make install
$ beamcore
Elixir 1.18+ Erlang/OTP 27+ Git

Beamcore

"Beamcore does not hide the machine. It gives the machine a voice, a memory, and a visible trail."
Execution path
User input flows to node A and node B, then into a generated function, local execution, and result