ZooWork

ZooWork is an AI agent delivery platform for domain experts and forward-deployed engineers: you package a role’s goals, company knowledge, SOPs, definition of done, and tool permissions into Agent Skills, build the agent in chat, and ship it to Slack, Microsoft Teams, WhatsApp, or into existing systems via API/SDK.

The runtime supplies planning, human approval gates, self-improvement from run feedback, and ZooData — a “any URL → structured JSON” layer advertised to cut LLM tokens by up to 75%. Models include OpenAI, Claude, Gemini, DeepSeek, Kimi, GLM, and others. A quick start is npx skills add SerendipityOneInc/zoowork-sdk-skills.

Highlights:

• Framed as deploying expertise as an on-the-job coworker, not another generic chatbot;
• Agents can be @-mentioned in chat apps or embedded beside internal tools;
• The company advertises a ~7-day assisted path to production for one real workflow;
• Site: https://zoowork.ai .

Why it matters

Most companies do not need another chat bot — they need role agents that can read/write real systems (replenishment, proposals, contract review) with approval gates. ZooWork packages skills, channel delivery, and runtime guardrails for that “works inside the IM / ops system you already have” path.

Limits

Results depend on how clearly you encode SOPs and permissions; wiring ERP/CRM still takes implementation work. Security claims (SSO, least privilege, no training on your data) should be checked against the contract. Industry figures on the site are illustrative, not promised ROI.

1. Create a project at https://zoowork.ai , or run npx skills add SerendipityOneInc/zoowork-sdk-skills ;
2. In Agent Builder, describe the role, knowledge, SOPs, and tool permissions in chat; preview and publish;
3. Connect the agent to Slack, Teams, or WhatsApp — or embed it via API/SDK — and require approval on writes;
4. Run it on real tasks and improve Skills from the feedback.