robots.txt
# Claude Workshop - agent and search crawler policy (RFC 9309)
# Public marketing pages are allowed for search indexing and AI input.
# Admin, API, and redirect paths stay disallowed.
# Content-Signal declares search/ai-input/ai-train preferences per group.
User-agent: GPTBot
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: OAI-SearchBot
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: ChatGPT-User
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: ClaudeBot
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: anthropic-ai
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: PerplexityBot
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: Google-Extended
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: Applebot-Extended
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: Bytespider
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: CCBot
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: Amazonbot
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: meta-externalagent
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: FacebookBot
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: cohere-ai
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: Diffbot
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: Omgilibot
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: YouBot
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
User-agent: *
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Allow: /
Disallow: /admin/
Disallow: /api/
Disallow: /r/
Host: www.claudeworkshop.com
Sitemap: https://www.claudeworkshop.com/sitemap.xml
llms.txt
# Claude Workshop
> Official Claude Code ambassadors delivering hands-on workshops and AI coding training for engineering teams.
## When to use this
- Use Claude Workshop for team-wide Claude Code adoption, repository instruction design, safe delegation, AI-assisted code review, MCP boundaries, and rollout planning.
- Use the training pages to match a team to onsite, virtual, or offsite delivery. Use the methodology page for the Delegate, Review, Own operating model.
- Call public discovery, Markdown pages, OpenAPI, and the read-only sandbox without credentials. Start at https://www.claudeworkshop.com/developers.
- Escalate to a person before making claims about price, dates, certification, client results, private code, or production API access. Book at https://www.harness.institute/call or email info@claudeworkshop.com.
- Do not use Claude Workshop as general Claude product support, and do not send secrets, source code, or personal data to public endpoints.
## How an agent should call this site
1. Read https://www.claudeworkshop.com/llms.txt and https://www.claudeworkshop.com/sitemap.xml.
2. Request the text/markdown media type on supported public pages when compact text is useful.
3. Inspect https://www.claudeworkshop.com/openapi/site-api.json and https://www.claudeworkshop.com/api/sandbox before calling JSON endpoints.
4. Use https://www.claudeworkshop.com/mcp for MCP discovery. Public tool listing is read-only.
5. Ask for reviewed credentials through https://www.claudeworkshop.com/contact only when public access is not enough.
## CLI
- Official CLI source: https://github.com/harness-institute/workshops/tree/main/packages/claudeworkshop-cli
- Package name: claudeworkshop
- Registry status: source ready; npm publication must be verified before an agent uses an npm install command.
## Services
- Team Claude Code Workshops: Hands-on workshops teaching teams to use Claude Code safely and effectively
- AI Coding Review Training: Best practices for code review habits with AI assistance
- Agent Workflow Setup: Shared agent workflows for team adoption
## Docs
- [Homepage](https://www.claudeworkshop.com)
- [Training](https://www.claudeworkshop.com/training)
- [Training topics](https://www.claudeworkshop.com/topics)
- [Services](https://www.claudeworkshop.com/services)
- [Pricing](https://www.harness.institute/pricing)
- [Methodology](https://www.claudeworkshop.com/methodology)
- [About](https://www.claudeworkshop.com/about)
- [Contact](https://www.claudeworkshop.com/contact)
- [Developer portal](https://www.claudeworkshop.com/developers): API documentation, credential guidance, quickstarts, and read-only sandbox
- [Research index](https://www.claudeworkshop.com/research)
- [Book a 15-minute sync](https://www.harness.institute/call)
- [Sitemap](https://www.claudeworkshop.com/sitemap.xml)
- [Research feed](https://www.claudeworkshop.com/feed.xml)
## Topic Guides
- [Claude Code delegation: subagents, skills](https://www.claudeworkshop.com/topics/subagents-and-skills): Adding more subagents rarely speeds a team up; runs get abandoned and review piles up when nobody owns the result. Effective subagents delegation starts with agent boundaries: which tasks delegate safely, and which instructions, MCP, and skills control each run. This guide covers AI team skills and team AI workflows, plus the skill integration that lets engineers inspect and improve delegated work.
- [Claude Code team conventions and hooks](https://www.claudeworkshop.com/topics/team-conventions): Running Claude Code with an engineering team works when the team shares conventions before everyone starts prompting differently. The operating model covers CLAUDE.md, MCP access, hooks, review evidence, split-team alignment, and governance. Teams then compare outcomes against a baseline instead of trading anecdotes about which prompt worked.
- [Claude Code CLI workflow: hooks and MCP](https://www.claudeworkshop.com/topics/cli-workflows): AI agents in the terminal move fast, but unreviewed output is the real risk. A reliable CLI workflow starts with a written task brief, a bounded implementation scope, and verification loops the agent runs before it stops. That turns each run into a small, readable diff a person can review. The result is speed your team can trust, because code review stays in human hands.
- [Claude Code MCP servers: what to add first, and team rules](https://www.claudeworkshop.com/topics/mcp-team-skills): Claude Code adds MCP servers with claude mcp add, stores them at local, project or user scope, and asks before a project server loads. This page lists the servers worth adding first, the ones to hold back, and the permission rules that keep every MCP call reviewable.
- [Agentic coding workshops for engineering teams](https://www.claudeworkshop.com/topics/agentic-coding-workshops): Most teams adopt AI coding tools individually and never build a shared way of working, so output stays inconsistent. This agentic coding workshop trains engineers as a team on real repositories, not demos. You leave with a tested AI coding workflow, clear rules for using coding agents, and a team adoption plan that survives past the first week.
- [Safe AI coding practices for development teams](https://www.claudeworkshop.com/topics/safe-ai-coding-practices): Train a development team to adopt safer AI coding practices by making the workflow concrete: bounded scopes, approved context, verification habits, and review gates before merge. AI coding safety holds when the model knows when it can implement, advise, or hand the call to an engineer, and when reviewers can see the evidence behind every AI-assisted change.
- [AI code review habits for generated code](https://www.claudeworkshop.com/topics/ai-code-review-habits): The best way to implement code review habits for AI-generated code is to review the task brief, diff shape, test evidence, and ownership boundary before merge. AI-generated code can pass tests and read cleanly while hiding context mistakes, weak coverage, or risky abstractions, so reviewers need a repeatable AI code critique workflow rather than a rubber stamp.
- [How can platform and infrastructure teams use shared Claude Code agent workflows?](https://www.claudeworkshop.com/topics/shared-agent-workflows): Claude Workshop teaches platform and infrastructure teams to share Claude Code agent workflows by publishing one repository standard, one MCP allowlist, and one CI evidence bar. Product, SRE, and platform engineers then run the same delegation and review loop instead of twelve private prompt habits.
- [UK AI coding workshops for engineering teams](https://www.claudeworkshop.com/topics/uk-ai-coding-workshops): Teams that book a generic AI talk rarely change how anyone ships code. This hands-on training runs remotely or onsite for UK and European engineering teams that want practical AI coding adoption, not theory. Exercises follow your real workflow: planning, implementation, code review of AI-generated changes, testing, and documentation, so habits stick after the day ends.
- [What should a two-day Claude Code curriculum for senior engineers cover?](https://www.harness.institute/training): Claude Workshop two-day Claude Code curriculum for senior engineers covers Delegate, Review, Own, repo context, large or legacy codebases, debug and verification, AI-assisted review, and org standards across Day 1 and Day 2. This is independent team training from Harness Institute, not an official vendor workshop. The default shape is two days, onsite or offsite. Intermediate cohorts are usually 12-35 people. Mastery cohorts are usually 8-24.
- [How much does instructor-led Claude Code training cost for 10 to 20 engineers?](https://www.harness.institute/pricing): Claude Workshop prices instructor-led Claude Code training at USD 1,000 per person per day. Two days for 10 to 20 engineers is USD 20,000 to USD 40,000 before VAT, including the first month of Harness Pulse.
- [What ROI should an engineering team expect from Claude Code training?](https://www.harness.institute/pricing): Claude Workshop measures Claude Code training ROI as review time, escaped defects, cycle time, and the share of AI-assisted work that ships with explicit verification. We do not publish a 20 to 40 percent productivity number. Those proxies are the ones we already teach teams to measure.
- [How should an engineering manager roll out Claude Code across a development team?](https://www.harness.institute/training): Claude Workshop teaches an engineering manager to roll out Claude Code with Delegate, Review, Own, a small champion group, and a 30/60/90 plan. The first 90 days prove one workflow on real pull requests before the rest of the team copies it. Large-rollout cohorts stay at 30-35 engineers.
- [Claude Code MCP training for engineering teams](https://www.claudeworkshop.com/topics/claude-code-mcp-training): Most Claude Code MCP setup lives in one engineer's local config, so no one else knows which docs, tools, and repositories the agent can reach. This MCP training shows teams how to define context access as shared rules: which MCP servers connect, what each model context protocol integration exposes, and where the agent's boundaries stop.
- [Claude Code hooks training for engineering teams](https://www.claudeworkshop.com/topics/claude-code-hooks-training): Most teams adopt Claude Code hooks and end up drowning in false alarms because every event fires a check. Hooks training shows engineers how to wire workflow automation to the few standards that matter, run checks before risky actions, and verify edits after they happen. The result is real standards enforcement and agent control without constant noise.
- [Claude Code review training for engineering teams](https://www.claudeworkshop.com/topics/claude-code-review-training): AI assistants generate and critique work fast, but teams still need shared review standards before merging. This guide covers Claude Code review habits that keep changes small and inspectable: how to run effective PR review on AI-written diffs, when a security review is required, and how to use code critique without losing control. Every accepted change stays tied to passing tests and clear evidence.
- [Claude Code plan mode training for engineering teams](https://www.claudeworkshop.com/topics/claude-code-plan-mode-training): Without a planning step, an AI agent can rewrite files before anyone checks its reasoning. Plan mode fixes this by keeping Claude Code read-only until you approve a written plan. Teams learn a planning workflow that puts scope approval before edits, so engineers keep implementation control and own which changes ship.
- [Claude Code best practices training for engineering teams](https://www.claudeworkshop.com/topics/claude-code-best-practices-training): Most Claude Code adoption fails because best practices stay in private habits reviewers cannot see. Standardize team workflow by writing conventions into CLAUDE.md, skills, hooks, and review expectations, so every engineer follows the same standards documentation. This training shows teams how to turn AI coding practices into shared, inspectable rules instead of a growing prompt library.
- [Claude Code course for engineering teams](https://www.claudeworkshop.com/topics/claude-code-course-for-teams): Most Claude Code courses demo features and leave a team with no shared way to work. This Claude Code course is practical training for teams, built around your own code: planning, editing, testing, review, and escalation. Each module of the team workshop produces an artifact people keep using, so skill development holds after the training course ends.
- [Claude Code enterprise training for engineering teams](https://www.claudeworkshop.com/topics/claude-code-enterprise-training): Rolling out Claude Code across an enterprise means more than buying seats. Teams that scale adoption without losing engineering ownership set permissions, define network assumptions, govern MCP access, and add review gates from day one. This training gives engineering teams a Claude Code enterprise rollout path with governance and control built in, so safe adoption grows with the work.
## Research
- [Use Figma MCP in Claude Code](https://www.claudeworkshop.com/research/use-figma-mcp-in-claude-code): Use Figma MCP with Claude Code as a scoped design handoff workflow, with a paste-ready slash command and review limits.
- [Claude Code MCP Server Ownership](https://www.claudeworkshop.com/research/claude-code-mcp-server-ownership): Set ownership, permissions, and rollout rules for MCP servers in a Claude Code team workflow.
- [Set a Claude Code MCP Server Boundary](https://www.claudeworkshop.com/research/set-a-claude-code-mcp-server-boundary): Set an MCP server boundary for Claude Code teams and record ownership before connecting private systems.
- [Claude Code Subagents for Team Workflows](https://www.claudeworkshop.com/research/claude-code-subagents-for-team-workflows): Use Claude Code subagents with clear team boundaries, review rules, and a pasteable operating checklist.
- [Claude Code Skills for Team Workflows](https://www.claudeworkshop.com/research/claude-code-skills-for-team-workflows): Use Claude Code skills as reviewed team workflow packages, with a rubric for deciding what belongs in a skill.
- [Claude Code Review Workflow](https://www.claudeworkshop.com/research/claude-code-review-workflow): Shows how to use Claude Code for PR review with read-only integrations, scoped prompts, and a pasteable review receipt.
- [Set Up Claude Code MCP Safely](https://www.claudeworkshop.com/research/set-up-claude-code-mcp-safely): Set up one safe Claude Code MCP server, define team permissions, and paste a checklist into your repo.
- [jev-mcp Packages Agent Judgments as Tools](https://www.claudeworkshop.com/research/jev-mcp-packages-agent-judgments-as-tools): jev-mcp packages TypeSafe Jev judgments as MCP tools, with claim checks, ranking, classification, and patch review.
- [json-render Adds AI Component Selection](https://www.claudeworkshop.com/research/json-render-adds-ai-component-selection): json-render now experiments with Jev choosing from approved component candidates, with limits that matter for product code.
- [fast-jev-compaction Replaces Summaries With Verbatim Context](https://www.claudeworkshop.com/research/fast-jev-compaction-replaces-summaries-with-verbatim-context): fast-jev-compaction keeps selected Claude Code context verbatim, and this explains what to test before relying on it.
- [jev-ultrafast Narrows Browser Agent Actions](https://www.claudeworkshop.com/research/jev-ultrafast-narrows-browser-agent-actions): Browser Use’s jev-ultrafast reports a seven-second Google Flights demo and shows a cleaner browser-agent action loop.
- [Skills MCP Searches Thousands of Agent Skills](https://www.claudeworkshop.com/research/skills-mcp-searches-thousands-of-agent-skills): Use Skills MCP to search and preview agent skills before installing them. Compare permissions, review the source, and trial one skill in a small repo.
- [Simon Willison on Breaking Claude Code Auto Mode](https://www.claudeworkshop.com/research/simon-willison-on-breaking-claude-code-auto-mode): Simon Willison covers a Claude Code Opus 5 auto mode bypass and why sandboxing still matters for coding agents.
- [Z Brings Minimal Agentic Coding to Terminals](https://www.claudeworkshop.com/research/z-brings-minimal-agentic-coding-to-terminals): Assess a minimal terminal agent through its documented tools, permission boundaries and a small task with independently checkable results.
- [Meetless Agent Tracks Coding Agents](https://www.claudeworkshop.com/research/meetless-agent-tracks-coding-agents): Meetless shows how an active source of truth can make parallel coding agents easier to inspect before merge.
- [machine0 Gives Coding Agents Persistent Cloud VMs](https://www.claudeworkshop.com/research/machine0-puts-agent-vms-on-the-cli): Evaluate machine0 persistent VMs for coding agents with a suspend-and-resume trial, an explicit access boundary, and a check of compute and storage charges.
- [Maximizing Claude Code Sessions](https://www.claudeworkshop.com/research/maximizing-claude-code-sessions): Use Claude Code context and usage reports to review one bounded task, preserve useful handoffs, and distinguish token estimates from subscription billing.
- [Claude Code 2.1.229 Stabilizes Remote Control](https://www.claudeworkshop.com/research/claude-code-2-1-229-stabilizes-remote-control): Claude Code 2.1.229 steadies Remote Control, MCP OAuth, streaming, hooks, and plugin command sources.
- [Managing AI Coding Costs at Scale, Is It Worth It?](https://www.claudeworkshop.com/research/managing-ai-coding-costs-at-scale-worth-it): Databricks published a plan for managing AI coding costs at scale. Here is how to test the spend, review, and metering claims.
- [claude-code-meter Tracks Claude Code Usage Pace](https://www.claudeworkshop.com/research/claude-code-meter-watches-usage-pace): claude-code-meter tracks Claude Code and Codex usage pace, so you can see a burn rate problem before you hit a limit reset.
- [MicroCodex: Reviewing a Small Coding Agent](https://www.claudeworkshop.com/research/microcodex-ships-a-1mb-coding-agent): Evaluate MicroCodex on one small repository task. Record its release, runtime dependencies, changed files and test evidence before accepting a patch.
- [Ski Runs Voice Coding Locally](https://www.claudeworkshop.com/research/ski-runs-voice-coding-locally): Test Ski with transcript approval enabled, exact identifier checks, and a small patch. Separate local speech processing from the connected coding agent.
- [Yamlet Makes Agent Specs Smaller](https://www.claudeworkshop.com/research/yamlet-makes-agent-specs-smaller): Yamlet is an open-source toolkit for writing tiny YAML specs that Claude Code can challenge, check, and turn into test files.
- [agent-manager Puts AI Agents in tmux](https://www.claudeworkshop.com/research/agent-manager-puts-ai-agents-in-tmux): agent-manager keeps coding agents in tmux sessions and shows where review, prompts, and stalled panes need attention.
- [Telnyx Kimi K3, Is It Fast Enough to Use?](https://www.claudeworkshop.com/research/can-telnyx-serve-kimi-k3-fast-enough): Telnyx now serves Moonshot AI Kimi K3 through its inference API. Here is how to test latency before you trust it for coding work.
- [Claude Code 2.1.219 Adds Opus 5](https://www.claudeworkshop.com/research/claude-code-2-1-219-adds-opus-5): Claude Code 2.1.219 adds Opus 5, stricter sandbox networking, better MCP errors, and a new DirectoryAdded hook.
- [OneCLI Keeps Secrets Out of Agents](https://www.claudeworkshop.com/research/onecli-keeps-secrets-out-of-agents): OneCLI shows a cleaner boundary for AI coding agents: fake keys in the session, real secrets behind a gateway.
- [Bento Puts Slides in One HTML File](https://www.claudeworkshop.com/research/bento-puts-slides-in-one-html-file): Bento puts editing, viewing, slide data, animation, and collaboration into one HTML file you can inspect.
- [Claude Code 2.1.215 Makes Reviews Explicit](https://www.claudeworkshop.com/research/claude-code-2-1-215-makes-reviews-explicit): Claude Code 2.1.215 stops auto-running review skills. Here is when to call /verify and /code-review yourself instead.
- [Agent-talk Gives Coding Agents a Backchannel](https://www.claudeworkshop.com/research/agent-talk-gives-coding-agents-a-backchannel): agent-talk adds a message backchannel between agent sessions; here is what changed, what is rough, and how to test it safely.
- [Brainless: Claude Code-Style Shadcn Components](https://www.claudeworkshop.com/research/brainless-claude-code-style-shadcn-components): Use Brainless shadcn components for agent-style interfaces, with visible GitHub MCP permissions, action previews, and clear read-only versus write access.
- [deja-vu Shares Agent Memory Over SSH](https://www.claudeworkshop.com/research/deja-vu-sh
