Skip to main content

SWAMP CLUB

The right way to build with AI

curl -fsSL https://swamp-club.com/install.sh | sh
Why it's betterBuild AI systems from repeatable behaviors, rather than token-hungry agent-centric designs. Use your favorite harness, connect to any API, store and query your data, keep secrets safe in vaults, and see exactly what happened every time.
Start your quest

builders already crushing it

[01]Moments of greatness

366,307,481 automation events

All-time automation events

1,887

Repeatable behaviors shared

[02]Top operatives this week

Full leaderboard

[03]Most-used extensions

  1. 1@swamp/s3-datastore1,010,946pulls
  2. 2@swamp/gcs-datastore679,554pulls
  3. 3@stack72/system-extensions233,842pulls
  4. 4@swamp/software-factory210,896pulls
  5. 5@swamp/1password207,080pulls
Browse the registry

[04]Heard in the night

01 / 09

Our marketers know exactly who they want to reach, and they shouldn't need a DevOps engineer to act on it. Swamp lets us give that control back to them, safely. […] Work that used to take hours or days now takes minutes.

Adam Roberts, XeroRead the source

What makes us unique as an industry is that we don’t just learn how to scale and support one application - we’re always learning how to build and support hundreds of new ones. […] We can integrate with anything.

Doug Schaefer, American SoundRead the source

I used Swamp to launch our production SaaS platform at Bitbison by simply describing the architecture we wanted, and letting Swamp build it all for me. What would have taken weeks of effort, we handled in a day […]

Samy Al Bahra, BitbisonRead the source

I found Swamp a very nifty tool to add rigor to my AI workflows, to reduce my inference costs […]

Dieter PlaetinckRead the source

Swamp does the same [configuration management as Ansible], and with "AI" in the loop, it does it faster and with results I'm more confident in.

Jean-Philippe EvrardRead the source

Swamp is the entire backbone [of my coding-agent orchestrator], and the thing I keep appreciating is how little of it there is.

Will McGinnisRead the source

Swamp lets me test the hypotheses instead of just hoping.

Mat GretenRead the source

Thirteen swamp workflows now cover everything [my 11 n8n flows] did, plus a few things n8n couldn't easily do.

sock-lintRead the source

Swamp is at the core a reusable automation framework for agents. Think of it as "multiplayer agents with forever memory".

Blake IrvinRead the source

[05]Field reports

  1. Testing Blender Scenes Built by LLM Agents1d agoby @aaddrick
  2. Replacing the newsroom with one workflow | 💻Gerard Klijs🦀3d agoby @gklijs
  3. Wrangling AI Agents for Predictable Results4d agoby @mgreten
  4. The System is on the Other Side of the Glass4d agoby @mellens
  5. Create Starter Projects and Provision Servers for Craft CMS with AI5d agoby @craftquest
Read the full feed

  1. Step 1:Any harness

    Swamp makes any harness or agent better.

    • Claude Code
    • Codex
    • VS Code
    • Cursor
    • OpenCode
    • Pi
    • Kiro
    • Your agent

  2. Step 2:Your intent → repeatable behaviors

    Tell your harness what you need, and ask it to build it with Swamp. Swamp teaches your harness to break your problem down into reusable behaviors, making it easy to scale, refactor, and extend.

    > Use swamp to build an agent that takes a Datadog alert, inspects the cluster, triages it, and routes it to Slack.
    1. 01
      fetch-alert
      @webframp/datadog/monitors
    2. 02
      inspect-cluster
      @swamp/kubernetes
    3. 03
      triage
      @webframp/triage
    4. 04
      route-to-slack
      @mgreten/slack-blocks

  3. Step 3:The right model at the right time

    Call behaviors directly from your agent, or have your behaviors call out to any frontier model.

    • Anthropic
    • OpenAI
    • Gemini
    • Grok
    • Jev
    • Kimi
    • Qwen
    • GLM

  4. Step 4:Work with any API, database, or service

    Swamp first searches thousands of existing extensions. If the service you need isn't covered, or isn't public, don't worry: Swamp extends itself to handle it.

    cloud providers
    databases
    observability
    services
    and 1,887 more

  5. Step 5:Store and query structured data & artifacts

    Each behavior can write structured data, indexed for search, or store its artifacts directly. Data is versioned and tracks its lifetime, so you control how much (and how long) information sticks around.

    • schema
      z.object({monitor: z.string(),severity: z.enum([…]),assignee: z.string(),})
    • artifact
      name: cluster-dump.tgzsize: 48 MBstored: as-islifetime: 30d
    • query
      severity == "p1"DD-40218 checkout-5xxDD-40231 ingress-cert2 matches · 4ms

  6. Step 6:Secure credentials

    Swamp pulls your credentials from secure vaults, so secrets are injected only when needed and redacted from the output.

  7. Step 7:Run it where you need it

    Call your new behaviors from your harness, from an AI agent, on the command line, or as an API.

    • Harness
      > triage checkout-5xx
    • Agent
      tool: alert-triage
    • Slack
      /triage checkout-5xx
    • CLI
      swamp workflow run alert-triage
    • API
      POST /hooks/datadog
    workflow
    alert-triage
    • ✓ fetch-alert
    • ✓ inspect-cluster
    • ✓ triage
    • ✓ route-to-slack

  8. Step 8:Scale out

    Start on your laptop, then scale out to run and manage Swamp centrally across your organization, with Swamp servers and remote workers.

    • laptop · you
      swamp cli
    • laptop · bobo
      claude code
    • ci runner
      cron · ci
    server
    swamp serve
    • grants & groups
    • dashboard
    • datastore
    • webhooks
    • build-7f3a
      fleet=build
    • build-c21e
      fleet=build
    • gpu-0b94
      fleet=gpu

  9. Step 9:Observe & audit the results

    Every command and session is tracked, leaving an audit trail of everything that happens. Get full observability through OpenTelemetry and your SIEM.

    trace · alert-triage · 3.4s
    • swamp.workflow.run
    • step fetch-alert
    • step inspect-cluster
    • model.method
    • step triage
    • step route-to-slack
    last 24h
    runs
    1,284
    p95
    3.6s
    failed
    0.4%
    export to
    • OpenTelemetry
    • Datadog
    • Grafana
    • Splunk
    • any SIEM

What your peers are building

  • Cloud & DevOps
  • Software Factories
  • Code Review & Verification
  • SRE
  • Observability
  • FinOps
  • Agent Sandboxing
  • Reproducible Data Science
  • Sovereign Cloud
  • Network Management
  • Identity Management
  • Enterprise AV
  • Homelabs
  • Personal Productivity
  • Marketing Automation
  • Sales Automation
  • Anything you want to build with AI, Swamp makes better.

It's time to build

curl -fsSL https://swamp-club.com/install.sh | sh

By using Swamp you agree to its software license and extension registry terms.