Skip to content

Ark

A lightweight, high-performance Slack gateway for AWS Bedrock Agents.

Ark connects Slack conversations to an AWS Bedrock Agent using WebSocket — no public endpoint or ingress is needed. When a message is received, it forwards the text to the agent and posts the streamed response back to Slack.

Features

  • Direct messages — the bot responds in a flat conversation
  • Channel mentions — the bot responds in a thread
  • Conversation context — thread timestamps are used as Bedrock session IDs, so follow-up messages maintain context
  • File support — upload files in Slack and they're sent to the agent's code interpreter
  • Table rendering — markdown tables in agent responses are rendered as Slack Block Kit tables
  • Citation sources — knowledge base sources are displayed as a bulleted list
  • Analytics — every interaction is optionally published to Kinesis Firehose for downstream analysis
  • AWS credential chain — explicit keys, ECS task roles, or AWS CLI (SSO, assume-role)
  • Minimal footprint — single static binary, under 20 MB memory, near-zero CPU at idle

Quick start

# Install Crystal 1.10+
# Clone the repository
git clone https://github.com/crystal-autobot/ark.git
cd ark

# Install dependencies
shards install

# Configure
cp .env.example .env
# Edit .env with your Slack tokens, Bedrock agent ID, and AWS credentials

# Run
make build && bin/ark

See Getting started for the full setup guide.

How it works

Slack (Socket Mode WS) ──> Ark Gateway ──> AWS Bedrock Agent
                              │
                              └──> AWS Firehose (optional analytics)
  1. Ark connects to Slack via Socket Mode (WebSocket)
  2. Incoming messages are forwarded to the configured Bedrock Agent via InvokeAgent
  3. The streamed response is converted from markdown to Slack's mrkdwn format and posted back
  4. Each interaction is optionally logged to Kinesis Firehose as newline-delimited JSON

Built with