Agent Catalog

General-purpose AI agent for coding, engineering, and build-it-yourself workflows.

What it does

OpenClaw is a production-grade general-purpose AI agent. It runs as a persistent worker on its own dedicated cloud server, maintains state across sessions, and can be wired up to any tool, API, or workflow you give it. Use it for software engineering, research, internal tooling, or any task that benefits from a flexible AI worker rather than a single-job specialist.

Who it's for

Developers, internal tools teams, solo builders, and founders shipping products. Anyone whose work involves defining the task, the tools, and the success criteria themselves. If you've looked at specialist agents and thought “none of these fit my job”, OpenClaw is the answer.

How it works

When you provision OpenClaw, Agent Launch spins up a dedicated cloud instance and configures the agent on it. You connect to it through the portal — terminal access, file system, and persistent context. From there, you direct it like you'd direct any technical contractor: give it a task, the tools it can use, and the constraints. It executes against those autonomously until the task is done or it needs you.

Integrations and surface area

• Webhooks in and out for triggering and reporting
• API access to common services your workflow needs
• Direct file system and git access on its own instance
• Custom CLI tooling configurable per project

Status

Live now. Available on every Agent Launch plan. Provisions on its own dedicated cloud instance in approximately 4 minutes.

AI customer support specialist for websites. Answers visitors 24/7 from your own content.

What it does

Support Pilot is a purpose-built customer support agent that embeds on your website as a chat widget. The initial crawl pulls your policies, FAQs, and product pages into a knowledge base — but that's only the starting point. From the dashboard, you can keep adding to what the agent knows: the user manual for each product, internal answer scripts, troubleshooting guides, vendor warranty documents, anything that doesn't live on your public website. The agent answers customer questions 24/7 using everything in the knowledge base, citing its source for every answer. When the agent isn't confident, it escalates the conversation to your team in the channel of your choice.

Who it's for

Small and mid-sized businesses with a website that receives customer questions. Especially well-suited for Shopify storefronts thanks to live product and order data lookups, but works on any site with content the agent can crawl. If your team answers the same five questions every morning — where's my order, do you ship to my country, what's your return policy — Support Pilot is built to take that off your plate.

How it works

Onboarding takes about ten minutes. You enter your site URL, the agent crawls your pages and builds an embedded knowledge base, and you paste a one-line snippet into your site theme to render the widget. From there, visitors ask questions in the widget and the agent answers using retrieval-augmented generation over your content. Confidence is scored on every answer; below the threshold, the conversation is handed off to your team.

Teaching the agent what it doesn't already know

The site crawl is the floor, not the ceiling. The Support Pilot dashboard includes a knowledge base editor where you keep extending what the agent knows about your business. The most common addition is a user manual for each product you sell — the document the manufacturer ships with the box, the PDF on the vendor's site, the troubleshooting guide your team wrote internally — uploaded once and now answerable by the agent forever. Same goes for return-policy edge cases, warranty procedures, internal answer scripts, vendor SLAs, anything your team knows that customers ask about but isn't on your public website. Each article you add becomes part of the same retrieval pipeline; the agent uses it just like crawled content and cites it as the source whenever it's the best answer.

When you find an answer the agent got wrong — usually surfaced through the conversation log — fix it once in the knowledge base, and every future customer with the same question gets the corrected answer.

Integrations and surface area

• Knowledge base sources: initial site crawl, plus dashboard-managed uploads — product user manuals, PDFs, internal procedures, custom-written articles. Add to or correct the KB any time.
• Shopify: live product availability and order status lookups (with order-by-email verification)
• Escalation channels: email, Slack, Microsoft Teams
• Help desks: Zendesk, Freshdesk, HubSpot, Salesforce, Jira
• Embed: one-line script on any page, works alongside existing chat or replaces it

Status

Live now. Provisions on its own dedicated cloud instance in approximately 4 minutes. Live in your site widget within about ten minutes end-to-end.

Full Support Pilot page →

General autonomy for teams that define the job themselves.

What it does

Agent Zero is a blank-slate autonomous agent that takes a task specification, a set of tools, and success criteria, and works on the job until it's done or it surfaces a question. Where OpenClaw is the technical builder's default, Agent Zero is the job-runner's default — closer to giving an autonomous task to a contractor than directing it line by line.

Who it's for

Teams comfortable defining the work themselves but who want the agent to drive execution. Often the right pick for ongoing operational tasks: monitoring something, triaging something, processing inbound work where each item is a self-contained job.

Status

In development. No committed release date. The order we build agents in is decided by customer demand — whichever agent the most businesses ask for goes to the front of the line.

Voice agent that answers your business phone line.

What it does

A voice agent that picks up your business phone line, handles common questions about hours, services, and appointment availability, books appointments directly into your calendar, and routes calls that need a human to the right person. Answers calls after hours and at lunch so customers stop hitting voicemail.

Who it's for

Local service businesses — clinics, contractors, salons, dental offices, accounting firms, shops, anyone where front-desk phone coverage is part of the customer experience. Especially valuable for businesses where missed calls turn into lost customers, and where a human receptionist is either too expensive or only covers a portion of the day.

How it works

You forward your existing business line to the agent's number. Calls come in, the agent answers in your business's voice, has a real conversation with the caller, and either resolves the call (answering a question, booking an appointment) or routes it to a human. Calendar integration handles the booking flow. Urgent calls — defined by you — get escalated immediately.

Status

In development. No committed release date. The order we build agents in is decided by customer demand.

Drafts responses to Google and Yelp reviews in your business's voice.

What it does

Reads new reviews on Google Business Profile and Yelp, drafts a reply that sounds like the owner wrote it, and either posts automatically or queues for human approval. Tone, common situations, and the actual voice of the business are all configured during onboarding so replies don't read like a template.

Who it's for

Single-location small businesses where the brand voice is the owner's voice — restaurants, salons, contractors, professional services. Businesses where letting a review sit unanswered for two weeks tells future customers more than any marketing copy will. Not built for multi-location chains where voice is centrally managed by a marketing team.

How it works

Connect your Google Business Profile and Yelp accounts during onboarding. The agent monitors for new reviews, drafts a reply in your configured voice, and either auto-posts (you choose the rules — e.g., auto-post for 4 and 5 stars, queue for human review on 1 to 3 stars) or surfaces the draft for you to approve in one click.

Status

In development. No committed release date. The order we build agents in is decided by customer demand.

Private search agent across your files, wikis, chats, and shared drives.

What it does

Indexes everything your team writes and stores — files, wikis, chat history, shared drives — and answers questions like a senior employee who's been at the company for ten years. The thing your team uses to stop asking “where is that document?” Internal-only, with full source citations on every answer.

Who it's for

Teams drowning in scattered knowledge across Slack threads from years ago, Google Drive folders nobody can navigate, and Notion workspaces with three competing sources of truth. Companies where new hires take months to get productive, where leadership re-explains the same things every quarter, and where the answer to a question always exists somewhere — nobody just remembers where.

How it works

Connect your data sources during onboarding — Google Drive, Slack, Notion, file shares, internal wikis. The agent indexes the content on your dedicated instance — your data does not leave your environment. Team members ask questions through a chat interface; the agent retrieves relevant content, generates an answer, and cites the source documents so the user can verify and dig deeper.

Status

In development. No committed release date. The order we build agents in is decided by customer demand.

Research, content, and outbound work.

What it does

Hermes is a general-purpose agent tuned for research and content workflows. Web research, summarization, draft generation, light outreach automation — the kind of work a junior team member might handle, scaled to whatever volume your operation needs. Same dedicated-instance architecture as OpenClaw, configured for research and writing rather than engineering.

Who it's for

Marketers, content teams, small agencies, solo operators doing research and outbound. Anyone whose work involves taking a prompt or brief and producing a researched, written output — especially at volumes where doing it manually is the bottleneck.

Status

In development. No committed release date. The order we build agents in is decided by customer demand.