02.04AI
AI agents
AI agents from Hot Rocket Software, a team of software, website and digital marketing specialists based in the UK, with 19 years in IT. AI agents that can plan and carry out multi-step tasks using your tools, with permissions, approvals and full logs. Built for UK businesses.
02.04.0
What is AI agents?
An AI agent is a model that can take several steps on its own to finish a task: look something up, decide, call a tool, check the result, and carry on. We build agents with narrow, clearly-defined jobs, explicit permissions, approval steps for anything risky, and a full log of what they did and why. Building practical AI tools that help a team work better is part of what we do day to day: we start with one clear job and add more only once it is working.
02.04.1
Who is it for?
- Teams with multi-step tasks that follow a pattern but need judgement
- Businesses with tools exposed through APIs or MCP servers
- Products adding an 'assistant that does things' rather than just answers
02.04.2
From a small business to a large organisation
The same team takes on a single tool for a small business and long-running work for large companies and the public sector. Here is what each end looks like.
For a small business
A small assistant that does one job end to end, such as chasing overdue invoices, inside rules you set.
For large companies and the public sector
Agents that work across several systems with defined permissions, monitoring and a person signing off important actions.
- Multi-step agents with approval gates
- Agent permissions and audit logs
- Evaluation before and after release
- Agents connected through MCP
02.04.3
What you get
- 01A written agent brief: goal, tools, limits and escalation rules
- 02The agent, with tool access through APIs or MCP
- 03Least-privilege credentials and approval gates
- 04Traces of every run for review and debugging
- 05Evaluation tests for the tasks it is trusted with
02.04.4
How we approach it
Discover
Pick one task with a clear finish line.
Scope
Which tools, which permissions, when to ask a human.
Build
Agent loop, tools, guardrails and tracing.
Launch
Supervised runs first, then gradual autonomy.
Support
Review traces, extend tools, re-test on model updates.
02.04.5
Typical stack
02.04.6
Questions
What is the difference between an agent and an automation?
An automation follows a fixed path. An agent chooses its next step. Use an automation when the path is known, and an agent when the path varies but the goal is clear.
Are AI agents safe to run?
They are as safe as the permissions you give them. We give agents the narrowest tools that do the job, require approval for anything irreversible, and log everything.
Do agents need MCP?
Not necessarily, but MCP is the standard way to give agents tools, and it means the same tools work across different AI clients.
02.04.7
What could this look like for you?
Example ideas, not client case studies.
- Inbox sorter: Incoming email is sorted, tagged and given a draft reply for you to check.
T+99Next step
Talk it through with us
A free 30-minute first chat, with no obligation. Tell us what is getting in the way; we tell you honestly what we would do, roughly what it would take, and whether it is worth doing at all.