Live demos of systems I have built, the patterns behind them, fair profiles of the harnesses people build agents with, and my own research. Every entry says honestly whether it is in production, built in the lab, or studied.
Nine small demos based on systems I have built. Pick one and watch the work get done.
Northgate Commandcommand.northgate.exampleDone
A chief of staff AI for a managing director. A request is typed into the command bar: sort the overnight emails, reply to a supplier about a delay, and ask the finance director agent about Q4 cashflow. A live trace shows the router sending the work to specialist agents. Eleven emails are triaged and a call is added to today's actions. A reply to the supplier is drafted and waits for approval. The board pack is summarised. The finance director agent reports £412k projected cash at year end, 9 percent ahead of plan, and flags two invoices over 45 days. The executive approves the reply and it is sent.
Morning brief, Wednesday 11 November
Good morning, Claire
3 tasks running3 tasks done5 agents online
Today
0809101112131415161718Stand-upBoard prepLunch, Tom B.Board pack reviewDebrief08:12
Ready for you to read
Q4 board pack summaryKnowledge agent, just nowReport
Amber, needs approvalApproved and sentPersonal assistant
Reply to Halcyon Supplies, draft ready
"Thanks Mark, understood on the delay. Can you confirm the revised delivery date by Thursday so we can update the Leeds team?"
ApproveEdit
Your directors report hereFinance, operations and research, when you ask
FDFinance DirectorQ4 cashflow assessment08:12
£412kYear-end projection9% ahead of plan
JulDec
Q4 cash position is healthy: £412k projected at year end, 9 percent ahead of plan. One risk: two large invoices are over 45 days. I'd chase them this week.
AI agents work without being asked when something other than a person wakes them: a schedule, a regular heartbeat check or an incoming event. Each wake-up gives the agent a fixed job, a time budget and a way to report back, so it can chase, prepare and flag things before anyone thinks to ask.
AI agents remember in layers: short-term context is what the model can see in the current conversation, and long-term memory is stored outside the model and fetched back in when relevant. Vector stores find memories by meaning, graph stores by how people, records and events connect.
Self-improving agents get better by writing down what worked as reusable skills: short procedures, checklists or scripts they load next time a similar task comes up. The model does not change; the library of procedures around it does, ideally with a person reviewing each change.
Routing means one agent reads each request and hands it to the specialist best placed to handle it, such as finance, operations or sales. Each specialist has its own instructions, tools and permissions, which keeps answers sharper and access tighter than one agent trying to do everything.
Sub-agents are helper agents that a lead agent starts for one well-defined piece of a larger job, each with its own fresh context, instructions and tools. Several can run at once and report back short findings, which makes broad tasks faster, at the price of more model usage.
AI agents reach real systems through tools: named functions such as "find invoice" that the model asks to run, while ordinary code does the actual work. MCP, the Model Context Protocol, is an open standard for offering those tools and data to any AI application in the same way.
Human approval gates sort every action an AI agent can take into three levels: draft only, needs a person to approve, or allowed to run automatically. The agent prepares, a person signs off anything that matters, and the system refuses any action that does not match what was approved.
An AI agent can follow a task for days or weeks only if the task lives outside the agent, as a saved record of goal, progress and next step that survives restarts. The agent wakes, reads the record, does one step, writes it back and sleeps, so nothing depends on one long conversation.
You put an AI agent in email, WhatsApp, Slack or Teams by connecting it to each platform's official messaging interface, so people message it where they already work. One core agent sits behind thin adapters that handle each channel's rules, such as WhatsApp's 24-hour reply window.
You see what an AI agent did and why by recording a trace of every run, each model call, tool call and decision as a timed step, plus a separate audit log of every action that changed something and who approved it. Traces are for debugging; the audit log is the permanent record.
You make an AI agent safe by limiting what it can do, not by trusting what it says. Give it the narrowest tools and permissions, run risky work in a sandbox, require a person to approve consequential actions, and treat every email, document and web page it reads as possibly hostile.
You control AI agent costs by sending routine work to small, cheap models and saving large ones for hard steps, caching repeated prompt content, and putting hard budgets and caps on every key, agent and job. Clients can then pay providers directly on their own keys or use a managed allowance.
You test AI agents with evals, a set of realistic tasks with clear pass or fail checks that you run every time the agent, its instructions or its model changes. The best tasks come from real failures, and each one runs several times because agents do not behave the same way twice.
4 min read
Harness profiles8
Well-known agent harnesses and frameworks: what each one is and what it does best.
OpenClaw is an open-source, self-hosted AI assistant that runs on your own machine and answers you in the chat apps you already use, such as WhatsApp, Slack, Teams and Telegram. It is best for one person or a small trusted team who want an always-on assistant with scheduled jobs, skills and tools under their own control.
Hermes Agent is an open-source AI agent from Nous Research that runs on a server or your own machine, talks to you through the terminal or chat apps, and writes reusable skills from the work it does. It suits people who want a long-lived agent that remembers, runs scheduled jobs and improves with use, on any model provider.
DeepSeek Harness (dsh) is DeepSeek's open-source agent harness, in public preview, built on the idea that every feature is a plugin. It suits developers who want a self-hosted coding and task agent they can reshape, running DeepSeek models or any provider that speaks a supported API.
Claude Code is Anthropic's agentic coding tool that reads a codebase, edits files and runs commands from the terminal, IDE, desktop app or browser. The Claude Agent SDK packages the same agent loop and tools as a Python and TypeScript library, for building your own agents on Claude.
The OpenAI Agents SDK is an open-source library, in Python and TypeScript, for building agents that use tools, hand work to each other and pause for approval. It suits teams that want a light agent loop with tracing built in, rather than a full orchestration platform.
LangGraph is an open-source framework from LangChain for building stateful agents as graphs of steps, with saved checkpoints so a run can pause, wait for a person and resume. It is best for long-running, multi-step work where you want explicit control over every path.
CrewAI is an open-source Python framework for building teams of agents, each with a role, a goal and tools, that work through tasks together. It is best for work that splits naturally into roles, wrapped in event-driven Flows that keep the overall process under control.
Microsoft Agent Framework is Microsoft’s open-source framework for agents and multi-agent workflows in .NET and Python, and the direct successor to both AutoGen and Semantic Kernel. It is best for teams on Azure or .NET that want typed, checkpointed workflows with approvals and OpenTelemetry tracing.
4 min read
Experiments3
My own research systems, built to find out what is possible.
Aura is an open-source AI companion with a state of mind that carries from one moment to the next: emotions that obey physics, memory that remembers how things felt, and five language models thinking in layers.
Magent is a no-code platform for creating, customising and deploying voice and text agents for customer support, workflow automation and personal assistance. Agents answer from their own knowledge base, search the web, and listen and speak.
The Discord guardian is a moderation bot that times out unwanted links and answers questions in English and Bulgarian, using Google Gemini. Slash commands set what it watches, so a community runs itself without a moderator online.
Nothing matches that yet
Try a shorter word, or . If you have a question the Lab does not answer, ask me directly.