AgentBuff

Small businesses, freelancers and creators have the same work as a company, from order recaps and customer replies to research and reports, but no team to hand it to. AgentBuff gives them AI employees that take instructions in the chat apps they already use, connect to their work tools and deliver finished work around the clock, with no technical setup.

The problem it solves

A small business has the same work as a large one: orders to recap, customers to answer, content to plan, research to run and reports to send. What it lacks is people to hand that work to. AI tools exist, but most of them expect the owner to learn a new app, write prompts from scratch or wire up integrations, which is exactly the time a small team does not have.

AgentBuff starts from the other side. Instead of asking people to come to an AI tool, it puts AI employees inside the chat apps they already use every day, connects them to the tools the business already runs on, and lets them work around the clock.

How AgentBuff works

Each account gets its own team of AI employees. An owner can set up several agents, such as Sales, Finance, Content and Research, and each one keeps its own context, skills and channels. Instructions go in over chat, and finished work comes back the same way: order recaps, customer replies, research, reports and documents. Agents also run scheduled tasks, so recurring work keeps moving outside office hours.

Setup takes about fifteen minutes, once, in four guided steps in the browser, and new users can try it free for 14 days. After that, the owner talks to the agents the way they would talk to a colleague.

See it at work

A three-minute demo recorded for a hackathon: why AI is still hard for a small business to adopt, then AgentBuff at work, from a chat with an AI employee to the skills marketplace and a cashier app installed as a skill.

What I built as founder

I founded AgentBuff and lead the product and the platform, from the architecture to the onboarding a non-technical owner goes through.

• A multi-tenant platform with an isolated 24/7 workspace for every user, so one customer's data and agents never touch another's.
• No-code onboarding that takes an owner from sign-up to a working AI employee without writing a prompt.
• The orchestration layer over 25+ AI providers, so users choose their models and keep their costs under control.
• 170+ skills and 63 tools, plus MCP integrations for systems outside the catalogue.
• BuffHub, the marketplace where skills and first-party products are packaged, including KostCloud, a boarding-house management system that AI agents can run over MCP.
• An admin control plane for users, billing, agents and audit trails.

Results so far

Users come from small businesses, clinics, education and freelancing. AgentBuff calls itself Indonesia's first AI employee ecosystem: it brings the agentic engineering used for enterprise systems to a solo business owner's phone.

Plans, from solo to enterprise

Plans are set per account, and the individual plan already includes unlimited agents. OP Buff is the plan for individuals, with unlimited agents, channels and tools. Full Managed, a fully managed option, is coming soon. Guild Master is the enterprise tier: corporate, BUMN and government needs are served there together with Spead AI, the enterprise AI company I also founded.

How AgentBuff relates to Spead AI

The two companies apply one idea at two scales. Spead AI serves organisations that need AI inside their own infrastructure, with data that never leaves the building. AgentBuff serves the people who cannot hire a team, with AI employees that live in their chat apps. When an AgentBuff customer grows into enterprise requirements, the Guild Master tier and Spead AI take over.

Security and data

• Each user runs in an individual data container.
• Users can delete their own files and results at any time.
• Access for support is limited and logged.

What building it taught me

The hard part of AI for small businesses is not the model. It is meeting people where they already work and making every agent predictable enough to trust. That is why AgentBuff lives in chat, why each agent keeps a narrow role with its own context, and why users choose their model instead of being locked into one.