Early AccessEvery agent is free to connect — no card, no checkout. Paid agents are coming.
Filter by Industry (who it's for) and Discipline (what it does). Pick either, or both.
Every agent is scanned and human-reviewed, and your API keys are only sent to the hosts it declares.How it works
FindAgent is a cross-LLM marketplace of focused AI agents. You find an agent, connect it to the AI app you already use — Claude, ChatGPT, Gemini, Cursor, the FindAgent CLI, or any MCP client — and use it where you already work.
Browse across two axes: the Industry an agent serves (who it is for) and the Discipline it embodies (what it does). Pick either or both, then refine by item type, tags, and integrations, or search for what you need.
Open an agent and follow its steps: install it locally with the FindAgent CLI, or, for a hosted agent, add its MCP server URL as a connector to wire it into Claude, ChatGPT, Gemini, Cursor, or any MCP client.
Every agent is scanned and human-reviewed before it is published, and your API keys are only sent to the hosts an agent declares. Declarative agents ship no executable code.
Every industry and discipline
Flags the days your GA4 metrics deviate from their baseline — with the numbers computed, not guessed.
Decides what your ad data can actually support, and refuses the rest
Turns YouTube Analytics metrics into a deterministic performance narrative + recommendations
Decides whether a change can merge: diff review, test evidence, breaking-change and tech-debt scoring.
Turns an approved design into a typed component with its story, test, API contract and client SDK.
Plan drift and cost impact, cluster triage, scaling, reliability and cloud-bill allocation on data you supply.
Stops an infrastructure or dependency change that opens a hole.
A 0–100 site health score from GA4, per device, with the failing metrics and what to do about each.
Funnel conversion and cohort retention from your own event data — every number computed, none narrated in.
Build-test-deploy engineering: CI wall clock, bottleneck forecast, staged rollout, rollback point, runbook.
Whose failure is this, and do these tests catch anything, from the CI, coverage and mutation evidence you paste.
Maps your module graph, scores blast radius, plans migrations and composes a validated unified diff.
Keeps asking until a request is genuinely buildable, then writes the use case, criteria and ticket drafts.
Campaign brief to a publish-ready landing page with tracking, an A/B variant and an enforced consent gate.
Detects idle Snowflake warehouses and expensive queries with monthly savings estimates
Reads lifecycle email against the counts, not against a benchmark nobody can source
CAC, LTV and payback per channel with a budget shift — and the attribution model always stated, never assumed.
Scores every shipped feature and tells a discovery problem apart from a value problem — before you sunset it.
Scores brand chatter for crisis risk and ranks the posts worth replying to — by reach, not recency.
Deterministic Mixpanel/Amplitude retention and churn analysis with LLM-narrated insights
Fuses Intercom + Stripe signals into churn risk scores with Salesforce + Slack alerts
Runs a month-end close review from your ledger: checklist, exception scan and what is blocking sign-off.
Zendesk CSAT trend analysis with negative-theme mining and Slack decline alerts
Scores HubSpot deals by health and surfaces at-risk opportunities with AI follow-ups
Ranks backlink prospects with drafted outreach — and demotes the link farms instead of recommending them.
Ages your receivables, scores collection risk and drafts reminders that match how late each account is.
Finds which feature's decline precedes cancellation — and how many days of warning it gives you.
Deterministic AML rule engine over Postgres transactions with SAR draft generation
Ranks the queries you can actually win, and says which tool it did not need
Turns your domain research into a sourced brief — and flags any claim without a source as critical.
Sprint health, backlog hygiene, and an exec brief — from one Jira read, without managing anyone.
Status-page updates and matching ticket replies — no invented causes, no invented ETAs.
Reads the Figma file and writes the brief — including the states nobody designed.
Drafts the knowledge-base articles and macros your resolved tickets are already asking for.
A front door for legal requests — categorised, complete, and never advised on.
Positioning, messaging, FAQ, and a dated checklist — all from what actually shipped.