AWS S3
Connic listens for ObjectCreated events and runs your agent on each new file. Extract structured data from PDFs, caption and tag images, transcribe audio, without writing a Lambda to do it.
Overview
S3 already tells you when objects change: bucket event notifications fire on every upload. The S3 connector gives each connector a unique webhook URL that receives those events through an SNS HTTP subscription or Amazon EventBridge, filters them by key prefix and suffix, and dispatches an agent run with the file's bucket, key, size, and event metadata. Optionally it downloads the file content and includes it in the payload.
The usual way to do this is a Lambda function per workflow: write it, package it, version it, wire its permissions, watch its logs. Here the event goes straight into an agent run instead, with the processing logic living in the agent you already deployed.
How it works
Create the connector
Add an S3 connector from your agent's Connector Flow. Enter the bucket name, region, and AWS credentials, plus optional prefix/suffix filters and content settings. The connector generates a unique webhook URL.
Point bucket notifications at the webhook
Configure your bucket to send events to the webhook URL, either through an SNS HTTP subscription or Amazon EventBridge. The connector confirms the SNS subscription, verifies each message's AWS signature, and unwraps notification envelopes automatically.
Upload a file
Every matching ObjectCreated event triggers an agent run with the file metadata, and the file content too if Include Content is enabled.
What you can build
Patterns teams ship in production. No queues, workers, or schedulers to run.
PDF uploaded
s3://invoices-bucket/*Agent reads the PDF, extracts vendor, line items, amounts, and due date as JSON, and writes the row to your AP table — replacing a manual data-entry queue.
PDF uploaded
s3://invoices-bucket/*Agent reads the PDF, extracts vendor, line items, amounts, and due date as JSON, and writes the row to your AP table — replacing a manual data-entry queue.
Event filtering
By default the connector triggers on Object Created events: Put, Post, Copy, and CompleteMultipartUpload. Switch to All Events to also react to deletes and restores. Key Prefix and Key Suffix filters narrow the trigger to the files you care about, like uploads/ and .pdf; events that don't match are silently ignored, so a busy bucket doesn't mean a busy agent.
Each event that does match becomes a normal agent run, with full traces, token and cost tracking, and the same guardrails and approval rules as any other trigger.
File content and credentials
The agent always receives the bucket, key, size, etag, and event details. With Include Content enabled, the connector downloads the file and adds it to the payload: text and JSON decoded as UTF-8, binary files base64 encoded, up to a configurable max size between 1 and 100 MB (files above it are skipped). Content download is the only part that needs IAM permissions, scoped to s3:GetObject and s3:HeadObject on the bucket. For larger files, the agent can fetch directly from S3 using the bucket and key. See the payload formats and setup commands in the S3 docs.
Information
- Publisher
- By Connic
- Category
- Connectors
- Modes
- Inbound
- Documentation
- AWS S3 docs
Frequently Asked Questions
How do I run an AI agent when a file is uploaded to S3?
Does the agent get the file content or just metadata?
Do I need to write a Lambda function?
Bring the event source, payload shape, result destination, and any private-network or approval requirements. We will map AWS S3 to the right Connic connector mode, deployment path, and observability setup.
Talk to Sales