Agent platforms

Agent platform consultancy

Six agent platforms shipped or matured in 2026. We help you choose the one that fits your work, set it up with the right permissions and roll it out to your team.

  • Meta Muse

    Meta · Hosted personal agent

    Runs long tasks in its own virtual machine and connects to email, calendars and payments. Meta opened it to small businesses in late September 2026 with Shopify, Dropbox and Slack integrations.

    Fits: Small businesses that already sell and talk to customers through Meta's apps

    Read the guide

  • OpenAI Dots

    OpenAI · Hosted always-on agent

    Works toward goals in the background with its own cloud computer and browser and connects to thousands of apps. Available on OpenAI's Pro plan and some business plans.

    Fits: Teams already on ChatGPT business plans

    Read the guide

  • Claude Duet

    Anthropic · Hosted proactive agent, research preview

    Builds daily briefings and dashboards and suggests actions from your data, running background tasks on its own virtual machine. A waitlist-based research preview in the US since October 2026.

    Fits: Professionals whose work lives in documents, inboxes and calendars

    Read the guide

  • Hermes Agent

    Nous Research · Open source, self-hosted

    Runs on your own server, learns skills from experience and is reachable over Telegram, Discord, Slack or WhatsApp.

    Fits: Teams that need the agent and its data on their own infrastructure

    Read the guide

  • OpenClaw

    Self-hosted agent platform

    Industry-shaped setup, workflow design and rollout planning for high-trust teams such as law, accounting, insurance and recruiting.

    Fits: Operationally dense teams with strict client and document rules

    OpenClaw consultancy Read the guide

  • Pi Durable

    Earendil · Library for durable agents

    An open-source TypeScript library from the team behind Pi. It checkpoints agent state as work progresses and resumes after a crash, with storage in SQLite, JSONL or Cloudflare Durable Objects. Still marked experimental.

    Fits: Engineering teams building their own long-running agents

    Read the guide

How a rollout works

  1. 1

    Discovery

    Which work the agent should take over, which data it may touch and what must never happen

  2. 2

    Platform choice

    A short comparison of the platforms against your data, plans and permissions

  3. 3

    Pilot

    One team, one workflow, with clear limits and a review after two to four weeks

  4. 4

    Rollout

    Training, guardrails and monitoring for the wider team, with documentation and handover

Tell us about your project

Let's talk