Claude Code training
Claude Code training from Ignite AI is one practical day for developers and technical teams using Anthropic's coding agent. We run it in person at your office in London or elsewhere in the UK, or remote, for rooms of up to 10, and every exercise runs on your own codebase.
Installing Claude Code takes one command. Agreeing how a whole team uses it takes a day.
Left alone, each developer builds private habits: their own prompts, their own permission settings, their own idea of what is safe to let it touch. The repo carries none of it, so the next person starts from nothing.
This day puts those decisions into the repo, where the whole team can read them, review them and change them.
Developers, tech leads and technical teams who work in a codebase every day. One team, up to 10 people per room.
It works best when the room shares a repo, because what we build together is then useful to all of them on Monday.
Right for you if
Better elsewhere if
One day, in your repo. Every feature below is a documented part of Claude Code, and each one links to Anthropic's own page so your team can go back to the source.
Morning: set it up so it holds
Everyone installs Claude Code and signs in through your organisation's accounts, then starts a session in your repo and asks it to explain the codebase back. We cover the terminal, the IDE extensions and the desktop app, so each person uses the one that suits them.
Anthropic docs: Quickstart, Authentication
CLAUDE.md is the file Claude Code reads at the start of every session. We write yours together: build and test commands, conventions, architecture decisions, the folders nobody touches. It goes into the repo and through review like any other change.
Anthropic docs: How Claude remembers your project
Permission modes decide what Claude Code can do without asking. We agree allow and deny rules, put them in shared project settings checked into the repo, and show where managed settings let an organisation enforce them.
Then plan mode: Claude reads the code and proposes a plan before it edits anything. Larger changes should start there.
Anthropic docs: Configure permissions, Settings, Permission modes, including plan mode
Claude Code works directly with git: branches, commits with useful messages, pull requests. We run the full loop on a
real ticket from your backlog, review the branch with /code-review, and keep a person approving every merge.
Anthropic docs: Common workflows, Review locally with /code-review
Afternoon: make it repeatable
A job your team repeats becomes a skill: a file in the repo that anyone runs as a /command. Custom commands
now live inside skills, so we build one or two for your own work, such as a release checklist or the way you write tests.
Anthropic docs: Extend Claude with skills, Commands
Subagents are specialised helpers with their own instructions and their own context. We set one up for a narrow job, such as reviewing changes or running the tests, so the main session stays focused.
Anthropic docs: Create custom subagents
Hooks run your own shell commands at set points: format after every edit, run lint before a commit, block a change to a protected path. A rule in CLAUDE.md depends on Claude following it, and a hook runs every time.
Anthropic docs: Automate actions with hooks
The Model Context Protocol connects Claude Code to tools your team already uses, such as your issue tracker or your docs. We connect one, show where project servers are configured, and agree which servers you trust. Anthropic says it does not security-audit or manage any MCP server, so that choice sits with you.
Anthropic docs: Connect Claude Code to tools via MCP
Claude Code also runs without the interactive screen, with claude -p, from scripts and pipelines. We wire up
one example in GitHub Actions or GitLab CI/CD, such as reviewing a pull request or triaging an issue, and set exactly
which tools it may use there.
Anthropic docs: Run Claude Code programmatically, GitHub Actions, GitLab CI/CD
We close on what leaves the machine, how secrets stay out and who reviews before merge, all set out below. Then your red lines. What never goes near a model. Who signs off what. Written down before anyone leaves the room or the call.
Before the day we hold a 30 minute scoping call. We agree the repo, the people in the room and two or three real tasks from your backlog to work on.
On the day, everyone works on their own machine, in that repo, under your own Claude accounts. In a room we walk round and look over shoulders. On a remote day we drop into each person's shared screen.
Everything the team builds is a file in the repo: CLAUDE.md, shared settings, skills, subagents, hooks and MCP configuration. It goes through your normal review, so it stays when we go.
Written and reviewed by the people who will rely on it.
Allow and deny rules in project settings, in the repo.
Built for jobs your team repeats every week.
What never goes near a model, and who signs off what.
We answer these with the code in front of us, using Anthropic's own published positions, linked so you can check them.
Claude Code runs locally. To use the model it sends data over the network, and Anthropic's documentation says that data includes all user prompts and model outputs, encrypted in transit. Code Claude reads for a task becomes part of that exchange, so we teach the team to treat anything Claude can read as something that may leave the machine.
Anthropic docs: Data usage
On commercial terms (Team, Enterprise and the API), Anthropic says it does not train generative models on code or prompts sent to Claude Code unless the customer opts in, and standard retention is 30 days. Individual Pro and Max plans have a model training setting the user controls. We check which applies to your team on the scoping call.
We add deny rules so Claude's file tools cannot read files such as .env or a secrets folder, turn on the
sandbox for shell commands, and keep credentials out of CLAUDE.md and out of prompts.
Anthropic docs: Configure permissions, Sandboxing
Anthropic's security guidance is plain: you are responsible for reviewing proposed code and commands before approval. So a person reviews every pull request Claude Code opens, and branch protection stays on.
Anthropic docs: Security
Where you need the policy behind those rules written and enforced across the business, that's AI governance, a separate piece of work.
The day is taught by our team of seven Claude Certified Architects, each holding Anthropic's Claude Certified Architect, Foundations (CCA-F) certification.
CCA-F is Anthropic's architect exam, launched in 2026 through the Claude Partner Network. Its domains include Claude Code configuration and workflows, prompt engineering and structured output, context management, MCP and agentic architecture. Those are the subjects of this day.
Ignite AI Solutions is a Claude Certified Services Partner in Anthropic's Claude Partner Network. We take no commission on licences, from any vendor, at any tier.
We run the day in person at your office in London or elsewhere in the UK, or remote, with the same outline either way. Our UK office is at 66 Paul Street, London EC2A 4NA.
We cap rooms at 10 because the afternoon is hands on, and we can't usefully look over more than ten shoulders at once, in a room or on screen. Bigger teams book several rooms.
Then we come back at day 30. We look at what stuck and what did not, and we fix it. Booked before the day ends, in person or on a video call.
This day is quoted after the scoping call. Prices for the Claude training day and our other training are on the pricing page.
The Claude training day is for knowledge workers in operations, finance, marketing, sales and admin, and nobody writes code. The Claude Code day is for people who work in a codebase. It runs in your repo, in the terminal, IDE or desktop app, and covers CLAUDE.md, permissions, skills, hooks, MCP and CI.
Yes. Everyone needs Claude Code access, through a Claude plan that includes it or an Anthropic Console account. We tell you exactly what to buy on the scoping call, and we take no commission on licences.
Claude Code runs on your machines, and it sends prompts and model outputs to the model, encrypted in transit. On commercial terms Anthropic says it does not train on code or prompts sent to Claude Code unless you opt in. On the day we set deny rules so Claude cannot read the files you mark off limits.
Yes. We run the day in person at your office in London or elsewhere in the UK, or remote, with the same outline either way. Our UK office is at 66 Paul Street, London EC2A 4NA.
We quote it after a 30 minute scoping call, once we know the team size, the repo and what you want the day to cover. Ask for a quote and we usually reply the same day.
Claude Code training
Tell us how many developers, where they are and which repo you want to work in. We'll come back with dates and a quote, usually the same day.