SWAMPCLUB

Adaptive workflows for AI agents.

You've asked your Agent to orchestrate a complex operation — and it does it right, most of the time.

Swamp teaches your Agent to build and execute repeatable workflows, makes all the data they produce searchable, and enables your team to collaborate in a predictable way.

Learn how swamp works

curl -fsSL swamp.club/install.sh | sh

GitHub Discord

01Building Blocks

Models

Typed interfaces to anything. Zod schemas in, validated data out.

Workflows

Multi-step DAGs with parallel jobs and nested composition.

Data

Immutable, versioned artifacts from every execution.

Vaults

Encrypted secret storage. Referenced by expression.

Extensions

Your agent writes one — it's a full citizen immediately.

Reports

Structured analysis after every run. Markdown + JSON.

02What Changes

Without

With Swamp

One-off scripts that rot

Typed, composable models

Capabilities lost in chat

Extensions that persist

Scattered credentials

Encrypted vaults

Ephemeral logs

Versioned, queryable data

03How It Works

INIT

Skills auto-discovered. Your agent knows the framework.

$ swamp repo init

BUILD

Your agent creates models from docs and your instructions.

Inventory every Proxmox VM and flag anything without monitoring

COMPOSE & RUN

Wire models into workflows. Review the YAML. Every run stores versioned data.

$ swamp workflow run inventory

04Real Usage

"Inventory every Proxmox VM and flag anything without monitoring"

Infra

"Remediate my Jellyfin library — find unidentified media and fix the metadata"

Media

"Rotate all API keys, update the vault, and notify Discord"

Security

SYS::ONLINE // open source // local-first