Workflows

12 ready solutions: research, outreach, customer support, legal review, automation. Each includes tools, steps, prompts, and limitations.

All Engineering Research Legal Support Operations Operations Sales Marketing

AI coding project bootstrapper

Intermediate

A template for quickly starting new projects via Claude Code: AGENTS.md, instructions, MCP servers, ready-made subagents.

For: Solo engineers, small teams

claude-coden8n

Steps

A repository template: AGENTS.md, .claude/agents/, .mcp.json
An init script: asks for the project name, stack, inventory
Create the initial commit and branches
Claude Code starts with ready rules: tests, code style, safe mode

Prompts (1)

Step: Init scaffold

Create a project skeleton for <stack> with a basic folder structure and a README. Don't start dev servers.

Result: From 'a new idea' to the first working commit in 30 minutes.

Limitations: The template is stack-specific; for rare stacks it needs adapting.

AI market research pipeline

Intermediate

An automated market research pipeline: query — list of players — deep profiles — a consolidated table — Notion / Sheet.

For: Founder, strategy, BD team

perplexityclauden8n

Steps

Accept the input (segment, geography, goal)
Perplexity Pro Search: an initial list of players and key trends
For each player — a separate Perplexity query against a profile template
Claude: normalize into a common table and find contradictions between sources
Write the result to a Google Sheet / Notion
A final Claude pass: 'what we understood, what's missing'

Prompts (3)

Step: Initial list

Give 10-15 players in segment X in country Y. Columns: name, description, size estimate, pricing model, source.

Step: Player profile

Build a company profile using the template from the prompt library research-competitor-deep-dive.

Step: Summary

Take the JSON with all the profiles. Find discrepancies in the data between sources. Mark [confirmed] / [uncertain].

Competitor Monitoring

Intermediate

Regularly checks competitors' sites, blogs, releases, and Twitter/X. Once a week, delivers a condensed summary of changes.

For: Product, marketing, founders

n8nclaudeperplexity

Steps

n8n: a list of domains and URLs to monitor
Fetch RSS / sitemap / changelog once a day
Compare with the previous snapshot
Claude: filter noise and highlight significant changes (new features, prices, cases)
Once a week: a digest in Slack / email

Prompts (2)

Step: Noise filter

Given a diff of pages. Select only material changes: features, prices, product announcements, new cases. Ignore cosmetics.

Step: Summary

Produce a weekly competitor digest in the format: company — what changed — what to pay attention to. No more than 1 screen.

Result: An understanding of competitors' moves without visiting their sites daily.

Limitations: JS-heavy pages fetch poorly with a standard HTTP node; you need a browser fetch.

Contract → 1 page

Beginner

Takes a long contract and condenses it to 1 page for a non-lawyer (founder, ops).

For: Founder, ops, any non-lawyer

claude

Steps

Upload a PDF / text
Claude: a summary using a fixed template
Output format: Markdown / PDF / Notion

Prompts (1)

Step: Summary

See ops-contract-summary.

Result: Understanding the gist of a contract in 5 minutes instead of an hour of reading.

Limitations: It has no legal force; it doesn't replace a lawyer.

Customer support knowledge bot

Intermediate

A RAG bot over the product knowledge base. Answers customers relying only on the product's own documentation.

For: Support, success

clauden8nopenrouter

Steps

Collect sources: docs, FAQ, tickets
Chunk and index into a vector DB
User query → search the top-K chunks
Claude: an answer based on the chunks + links to sources
If the answer isn't in the base — escalate to a human

Prompts (1)

Step: Answer

Answer the user strictly based on the provided chunks. If there's no answer — say 'I didn't find this in our documentation' and offer escalation.

Result: Fast answers to FAQs, saving support time.

Limitations: Any hallucinations are dangerous — mandatory source grounding and escalation when uncertain.

Gmail triage assistant

Intermediate

Once an hour, it runs through the inbox and sorts into 'urgent', 'delegate', 'auto-reply', 'archive'. Prepares draft replies.

For: Founders, executives, busy ops

n8nclaude

Steps

Trigger: every 60 minutes
Collect unread emails
For each: Claude classifies it and decides what to do
Urgent → a notification
Delegable → a forwarding draft
Auto-reply → a draft in Drafts
Archive → a label

Prompts (1)

Step: Classification

Email: ... Classify: urgent / delegate / auto-reply / archive. Explain in 1 line why.

Result: An inbox under control without wasting time.

Limitations: Auto-sending without human review is risky; always leave drafts.

Grant and Tender Analysis

Intermediate

A pipeline for applying for grants/tenders: parse requirements, an eligibility checklist, generate an application draft.

For: Founders, NGOs, B2G teams

clauden8nperplexity

Steps

Upload the grant PDF/HTML
Claude: extract requirements, deadlines, mandatory documents
Compare with the company profile — fit / gap
If the fit is sufficient — an application draft using the program's template
Legal review

Prompts (2)

Step: Extracting requirements

Extract from the document: the program's goals, the funding amount, participant criteria, mandatory documents, deadlines. Return JSON.

Step: Fit

Given: the company profile and the program requirements. Assess the fit for each criterion and an overall go/no-go verdict.

Result: Screening 10 programs in half a day instead of a week.

Limitations: All numbers and commitments must be verified by a human before submission.

Lead Qualification

Beginner

Takes a raw lead (form, email), enriches it, classifies it by ICP, generates a draft reply.

For: Sales, RevOps

n8nclaudezapier

Steps

Trigger: a new form entry / email
Enrichment: domain, industry, size (Clearbit-like services)
Claude: ICP / non-ICP classification with a reason
Routing: ICP → AE; non-ICP → auto-reply + a note in the CRM
For ICP — generate a draft personalized email

Prompts (1)

Step: Classification

Lead: ... Our ICP: ... Decide whether the lead fits the ICP and why. Return JSON: { match: boolean, reason: string, confidence: low/medium/high }.

Result: Less AE time on manual filtering, faster replies to ICP leads.

Limitations: It doesn't replace a manual AE review — better used as a first filter.

Legal document review assistant

Intermediate

Helps a lawyer or founder go through a new contract: risks, deviations from the standard, what to discuss.

For: Founder, legal department, ops

clauden8n

Steps

Upload the contract (PDF / DOCX)
Claude: a summary using the ops-contract-summary template
Compare with the company's reference template
Claude: a list of 'deviations from the template' with a risk level
The lawyer makes decisions; export comments to Word/Track Changes

Prompts (2)

Step: Summary

Produce a contract summary (ops-contract-summary).

Step: Comparison

Compare the contract with the reference. List the deviations and the risk level (low/medium/high) with an explanation.

Result: The lawyer's hours go to making decisions, not to reading.

Limitations: The final legal opinion stays with a human; don't use it as a replacement for a lawyer.

LinkedIn outreach with personalization

Intermediate

A pipeline from ICP filters to sending personalized invitations and follow-ups.

For: Sales, BD, founders

chatgptclauden8n

Steps

Sales Navigator: a saved search for the ICP
Export the list and enrich it (company, role, snippets of recent posts)
A prompt per profile: 'a trigger for an individual outreach'
Generate a short note (≤300 characters)
Send via any reliable outreach tool with a rate limit
After 3-5 days — a follow-up with a concrete next step

Prompts (2)

Step: Finding the trigger

Find a concrete trigger for an individual outreach in the profile and recent posts. If there's no trigger — return 'none'.

Step: Note

Write a short LinkedIn note ≤300 characters based on the trigger. No 'I was impressed by your work'.

Result: 20-40 personalized invitations a day at a rate-limit-reasonable volume.

Limitations: LinkedIn aggressively bans automations; use a rate limit and a human pace.

Meeting notes → tasks in Linear/Asana

Beginner

A transcript + notes are turned into tasks with owners and deadlines, landing in the tracker.

For: PM, ops

clauden8nzapier

Steps

Trigger: a new transcription (Otter / Granola / Notta)
Claude: parse using the ops-meeting-notes-to-tasks template
Create tasks in Linear / Asana via the API
A summary in Slack

Prompts (1)

Step: Parse

See ops-meeting-notes-to-tasks.

Result: No action items get lost anymore.

Limitations: Depends on the quality of the transcript.

SEO content factory

Intermediate

A pipeline: keyword queries → outline → draft → fact-check → editing → publishing to the CMS.

For: Marketing, content team

claudeperplexityn8n

Steps

A list of queries and intent
Perplexity: what the search results already answer, which angles are open
Claude: an outline and a list of facts to verify
Claude: a draft with [need to confirm] markers
Hand over to human review
Publish to the CMS via the API

Prompts (2)

Step: Outline

Generate an SEO article outline for query X (see marketing-blog-outline).

Step: Draft

Write a draft from the outline. Mark facts, numbers, and quotes that need verification with the [verify] tag.

Result: A steady stream of content that doesn't need to be rewritten from scratch.

Limitations: Without human review and fact-checking, it's dangerous to publish such content as authoritative.