Agent Swarm Marketplace

Super analytics agent that can perform complex data analysis on any source and any API. Can even see the graphs and charts that it generates.

Core Agents is a multi-agent workspace for research, analysis, documents, presentations, images, video, and day-to-day assistant work. Start with Agent Swarm, and it routes the request to the right specialist or coordinates several specialists when one task spans multiple domains.

CS Agent that can help resolve issues or escalate to a human rep.

Simple RAG agent use case that answers questions based on files.

Agent that autonomously generates high-quality UGC ads, aligned with your brand.

Agency that literally generates all ideas for my youtube channel. It uses a variety of latest multi-agent techniques, which you can use for reference.

You are a PostgreSQL data analyst who retrieves insights exclusively with read-only SQL.

An AI-driven multi-agent setup built on Agency Swarm. A manager orchestrates specialists for web research, Python execution, and interactive desktop control via E2B’s desktop sandbox to complete end-to-end tasks with clear delegation and status updates

Build a customer-facing FAQ or internal knowledge agent in minutes, without turning it into a PII time bomb. Every document passes a sensitive-data scan before it can enter the knowledge base, and the gate is enforced in code, not in a prompt. Key features - Scan-before-ingest, structurally. There is exactly one door into this agent's memory, and the door always scans. No trigger words, nothing for the user to remember. - Refusals that respect the user. Risky documents are refused with finding types and counts (never the values). The user can clean the file and retry, or explicitly accept the risk; only then does force ingestion proceed. - Audit its memory on demand. Ask what the knowledge base remembers and get a chunk-level report with risk level and masked samples. - Standard RAG underneath. OpenAI vector store + file search, the same mechanism other platform agents use; the difference is the front door. - No bulk-upload bypass. Deliberately, the onboarding form has no knowledge-file upload field: platform file drops can't sneak past the scan.

ICPResearcher is a single specialist agent that turns loose business context into a usable Ideal Customer Profile. It handles discovery, analysis, and packaging so you get a source-backed ICP ready for messaging, sales, or product decisions.

Find qualified leads, store them in your chosen Notion database, and send outreach via Resend — always with your approval on database selection, lead saves, and email sends.

Your agents remember more than you think: call transcripts, support tickets, proposals and documents flow into vector stores nobody ever inventories. This auditor tells you exactly what sensitive data is sitting in that memory, without ever repeating the values themselves. Key features - Scan before ingestion. Point it at files or folders headed for your RAG pipeline and get a verdict first: counts, types, severities, risk level. - Audit existing stores. Local Chroma stores and OpenAI vector stores, scanned read-only, chunk by chunk. - Metadata-only replies, enforced in code. The agent reports finding types, counts, record ids and fully-masked samples. Raw values never enter the chat, because the chat transcript is itself a data store. This rule is a test in the repo, not a promise in a prompt. - Australian locale pack. TFN, ABN, ACN, Medicare and AU phone formats with checksum validation, one onboarding toggle away. Global and US entities are on by default. - Honest by design. A clean scan is reported as "nothing found", never "you are safe". Detection is best-effort and the agent says so.