For Technical Teams · Build & Extend

The Same Craft,
Infrastructure Handled

If your team already builds with Claude Code, you already know how to build on Ai1. Same open skill format, no proprietary DSL, and the servers, scheduling, deployment, and credential store are handled for you. Bring what you have, or take what we build.

A builder's holographic workspace with modular skill blocks snapping together

Same open standard you know

Skills Are Just SKILL.md Files

A skill is a plain markdown file with a little structure: a name, a description, and the instructions an agent reads and follows. It's the same format as Claude Code and Codex. There's no proprietary domain-specific language to learn, so your team is productive on day one.

  • Author in plain markdown, version it like code.
  • Reference scripts by path for anything that needs real execution.
  • Compose skills into recipes, repeatable procedures with QA built in.
# skills/publish-weekly-report/SKILL.md
name: publish-weekly-report
description: Compiles last week's metrics
  into a branded report and posts it.

## Steps
1. Pull metrics for the date range
2. Draft the summary in the brand voice
3. Render to HTML and publish
4. Post the link to the team channel
Modular glowing components and building blocks on a precise workbench surface.

Bring What You Have. Build What You Need.

Bring your existing agents

Have your current Claude Code or Claude Cowork agents export their skill files and related docs, zip the project folder, and hand it to your Ai1 agent. It imports everything automatically. Nothing to rewrite.

Build your own

Author new skills, compose them into recipes, and stand up new specialist agents, all under governance and human approval. The platform can grow its own capability surface, safely.

MCP-native

Live Tools and Your Own Knowledge

Mount Model Context Protocol servers for live tool access, and query your own organization's knowledge from the tools you already use. The brain isn't a walled garden. It's a standards-based layer you can reach from Claude Code, Claude Desktop, or any MCP-compatible client.

Real execution

Not a Sandbox

Agents run on real infrastructure with scheduling, deployment, and a logged credential store. They do actual work on a schedule and on demand, not a toy demo that forgets everything when the tab closes.

An expansive real production environment of infrastructure, room to run at scale.

Bring Your Own Model and Tokens

Connect the providers you already pay for. You pay them directly, with no markup and no lock to a single vendor. Mix and match by task.

Language models

Claude, plus OpenAI through its CLI and Codex, and Cursor. Route work to the right model for the job.

Images and video

Gemini for image and video generation, wired into your content workflows.

Voice and phone

ElevenLabs for voice agents and Twilio for real phone numbers.

No markup. You connect your own tokens and pay your provider directly. We don't sit in the middle of your model bill.

What It Connects To

A practical catalog, grouped by what you're trying to do. Connect with your own credentials, scoped per channel.

Productivity & docs

Google DriveGoogle SheetsGmailGoogle Docs

Analytics & marketing

Google AnalyticsSearch ConsoleGoogle Ads

Comms & channels

SlackEmailSMSVoice

Project & ops

AsanaTwilioDNS & domains

Catalog grows over time. If a tool exposes an API or an MCP server, it can be connected.

See the full Ai1 platform