Meet OpenPipal

Open-source visual AI Agent, built on Pi · No terminal required

Download for desktop View on GitHub

macOS & Windows · All downloads ↗

✦ General Agent

What should your Agent work on?

Example Agents Examples only · not preinstalled

Generated from conversations

＋

Create an Agent from a conversationKeep its agent.md, memory, skills, and task scope in one workspace.

Skills and tools

Extend the Agent with standard plugins, Skills, MCP servers, and CLI tools.

⌘Install Agent PluginStandard package: plugin.json + Skills + MCP

⌘

release-agent-kit v1.2.0Release checks, GitHub evidence, and one MCP server3 Skills · 1 MCP server

research-source-pack v0.8.4Source capture and evidence-first research2 Skills · 2 MCP servers

Weekly product intelligenceSucceededAgent · Product Research•Every Monday 09:00•Next: Aug 17

Release readiness watchRunningAgent · Release Review•Webhook•Started 2 min ago

Content source digestPausedAgent · Content Studio•Every day 18:00

Review this release against our acceptance gates and flag anything that blocks shipping.

Release Review

I split the work across two Subagents. You can inspect each step and the evidence as it arrives.

✦Subagentrelease-evidence✓via GLM-5

Steer the task or add context… SkillsFilesGLM-5↑

⌄▱Outputs

⌄✓Automation

⌄⌁Sources

package.jsonbuild.logCHANGELOG.md

01/06

Start in the app, not the terminal

Pick an Agent, connect the model service you use, and begin with one real job.

Click the app to explore

01 / Built on Pi

Keep the model at the center.
Complexity, only when needed.

The Pi Agent framework, with a visual interface. A minimal core, controlled context, and tools only when needed.

Capabilities can grow. Complexity does not have to be the default.

Real taskReview this release

Context3 relevant files

SkillLoaded on demand

ToolOnly when needed

Pi core Model Working with what matters

Finished resultProcess and evidence stay visible

Only what this task needs enters the loop.

02 / Dedicated Agents

For every job.
A dedicated Agent.

Dedicated Agent

Release Agent

Open foundation

ACP

MCP Agent Skills Agent Plugins

03 / ACP

Design it once.
Reuse it across your workflows.

The same Agent core can live in the app or inside the flow that needs it.

One Agent core OpenPipal

New escalation Response plan returned to the workflow

1 Agent · N workflows

Keep attention for judgment, creation, and decisions.

04 / Automation

Run it once.
Or keep it running.

Manual task

Release readiness review

Starts when you ask

05 / Enterprise

Deploy in your environment.
Make it your system.

Request an enterprise deployment

01Private deploymentYour infrastructure

02Models and dataYour boundaries

03Agent experienceYour workflows

Apache 2.0 · Local-first · macOS & Windows

Start with one real job.
Meet OpenPipal.

Connect your model. Your first dedicated Agent is minutes away.

Download for desktop Explore the source

macOS & Windows · All downloads ↗

Meet OpenPipal

What should your Agent work on?

My Agents

Skills and tools

Automation

Works

Keep the model at the center.
Complexity, only when needed.

For every job.
A dedicated Agent.

Release Agent

Design it once.
Reuse it across your workflows.

Run it once.
Or keep it running.

Release readiness review

Deploy in your environment.
Make it your system.

Start with one real job.
Meet OpenPipal.

Download OpenPipal

Meet OpenPipal Dedicated Agents for every job

What should your Agent work on?

Keep the model at the center.Complexity, only when needed.

For every job.A dedicated Agent.

Release Agent

Design it once.Reuse it across your workflows.

Run it once.Or keep it running.

Release readiness review

Deploy in your environment.Make it your system.

Start with one real job.Meet OpenPipal.

Meet OpenPipal

Keep the model at the center.
Complexity, only when needed.

For every job.
A dedicated Agent.

Design it once.
Reuse it across your workflows.

Run it once.
Or keep it running.

Deploy in your environment.
Make it your system.

Start with one real job.
Meet OpenPipal.