Open-source agent framework

Build agents that
run in Python.

OpenProgram gives you the runtime, Web UI, terminal UI, model connections, tools, skills, memory, and multi-agent collaboration. Install a ready-made harness or build an agent for your own workflow.

Install OpenProgram Explore documentation

OpenProgram Web UI

Run agents, inspect sessions, and review every call from one interface.

Start where you need

Use an agent or build your own.

Install an existing harness for a complete workflow, or use the same OpenProgram runtime to create an agent in ordinary Python.

openprogram programs install gui

GUI Agent

Operate desktop applications and GUI environments with visual grounding, action execution, and task verification.

View installation guide → openprogram programs install research

Research Agent

Run literature research, experiment, analysis, and paper-writing workflows from one agent harness.

View repository → openprogram programs install wiki

Wiki Agent

Organize notes, documents, and conversations into a maintainable Obsidian knowledge base.

View repository → Build your own

Your agent, ordinary Python

Use a decorated function for the task contract and call the model only where the workflow needs judgement.

@agentic_function
def triage(ticket: str, runtime=None):
    kind = runtime.exec(ticket,
        choices=["bug", "feature"])
    return handle(ticket, kind)

Read the harness guide →

Quick Start

Install, then run.

The first run guides you through connecting a model provider and choosing the Web UI or terminal interface.

Windows and advanced installation →

Included framework

Everything you need to run agents.

Use the complete runtime or adopt only the components required by your workflow. Configuration and execution remain available from Python, the Web UI, and the terminal.

providers

Model providers

Connect Anthropic, OpenAI, Gemini, and compatible backends with account and model configuration.

tools + mcp

Tools and MCP

Give agents local tools, browser actions, external services, and MCP server capabilities.

skills + memory

Skills and memory

Load reusable instructions and retain task knowledge across longer-running work.

multi-agent

Agent collaboration

Spawn agents, exchange messages, and isolate file-changing work in separate branches.

sessions

Sessions and branches

Keep conversations, calls, context, forks, and execution history inspectable.

web · tui · cli

Three interfaces

Run the same agents from the browser, terminal UI, or headless command line.

channels

Channel integrations

Connect agent workers to supported messaging channels without replacing the runtime.

programs

Installable harnesses

Install, manage, run, and publish Python harnesses through one package contract.

One runtime, several surfaces

Use the interface that fits the task.

The Web UI, terminal UI, and headless CLI share sessions, providers, tools, skills, memory, and the same execution DAG.

Web UITerminal UICLIProvidersMCPSkillsMemoryChannels

OpenProgram code interface showing an agentic Python workflow — **Agentic functions**Model calls and deterministic Python in one execution flow.

OpenProgram Web UI showing a conversation and execution DAG — **Web UI**Inspect sessions, branches, calls, and context.

OpenProgram terminal UI showing providers, tools, and available agents — **Terminal UI**The same sessions and runtime from a terminal.

Framework internals

Inspectable context, control, and events.

OpenProgram keeps agent execution inside Python while recording model calls, context dependencies, validation, and runtime events.

caller + reads

DAG Context

Every user turn, model call, and function call is a node. Dependency edges select the context used by branches, sub-agents, and cross-session messages.

Read the context documentation →

Python gates

Agentic Workflow

Python fixes required execution order, validates model decisions, retries failed checks, and combines agentic functions with ordinary tools.

Explore runtime features →

Event(type, payload, ts)

Event Infrastructure

A process-wide event bus carries typed activity from the agent loop, authentication, context, channels, and memory.

Inspect the event design →

Research

Build your next agent with OpenProgram.

Install the framework, connect a model provider, and start from an existing harness or your own Python workflow.

Installation guide Browse documentation