Summary

Swamp AI is a CLI tool designed to empower AI agents in creating operational workflows that are auditable, shareable, and accurate. It targets developers and power users seeking to automate complex tasks securely and transparently, with workflows reviewed locally before production deployment. The primary benefit is enhanced human oversight and control over AI-driven automation.

How It Works

Swamp employs a model-based approach, defining external systems (cloud resources, CLI tools, APIs) as "Models." These are configured via YAML "Definitions" that leverage CEL expressions for dynamic values and cross-model references. "Workflows" orchestrate model executions across parallel jobs and steps. All artifacts, data, and secrets are managed locally within a .swamp/ directory or through configurable datastores (S3, shared filesystem). The tool integrates AI coding assistants by providing them with specific "skills" to interact with the Swamp CLI, enabling agents to discover and utilize automation capabilities.

Quick Start & Requirements

Installation is performed via a shell script: curl -fsSL https://swamp.club/install.sh | sh. Initializing an AI agent workflow is done with swamp repo init (defaulting to Claude Code) or specific flags like --tool cursor. Development requires Deno. The tool automatically picks up environment variables (e.g., AWS credentials, SSH keys, kubeconfig) for local execution.

Highlighted Details

• AI Agent Native: Ships with first-class skills for Claude Code, Cursor, OpenCode, and Codex, enabling AI agents to discover and utilize Swamp's automation capabilities seamlessly.
• Local-First Security: All operations and sensitive credentials remain on the user's machine unless explicitly invoked via an API, significantly enhancing supply chain security.
• Flexible Datastores: Supports local filesystem, shared network drives, and S3 for centralized or collaborative data management and state sharing across machines.
• Issue-Driven Contributions: External contributions are managed via a strict issue-filing process, with maintainers implementing changes to ensure code quality and security, offering co-authorship credit to users.

Maintenance & Community

The project emphasizes an issue-driven contribution model, where external users file issues that are then triaged and implemented by the core team, with credit given to the issue filer. A Discord community is available for users.

Licensing & Compatibility

Swamp is licensed under the GNU Affero General Public License v3.0 with a specific Swamp Extension and Definition Exception. AGPLv3 is a strong copyleft license, which may impose obligations on derivative works and could impact integration with proprietary software.

Limitations & Caveats

Direct pull requests from external contributors are not accepted, a deliberate choice for supply chain security. The AGPLv3 license imposes significant obligations on derivative works, requiring them to be open-sourced under the same license, which can be a constraint for some commercial use cases.