OpenAI builds and deploys advanced AI models like GPT-4o for autonomous agents and workflows.
subs
$0.02Starting price
0Popularity
About subs
subs is a cloud-native agent harness designed to run unprivileged agent loops without system access. It operates either locally or as a client-server system, using MCP servers for tool integration. The tool supports local development via a CLI and can be deployed in hosted or self-hosted modes. Users configure agents through a TOML file specifying LLM providers, system prompts, and MCP connections. The engine executes agent turns, handling model calls, tool invocations, and sub-agents within a single turn. It integrates with Slack for direct messaging or channel mentions, supports MCP connectors for external tools, and allows customization via webhook workers. The system enables durable sessions with full event logging and AG-UI streaming to compatible clients.
Key features
- Unprivileged agent loop execution
- MCP server integration for tools
- Local and client-server deployment
- Slack integration for DMs and mentions
- Webhook worker customization
- Durable sessions with event logging
- AG-UI streaming support
- Multi-agent and sub-agent support
Use cases
- Building Slack bots for team collaboration
- Automating workflows with MCP-connected tools
- Deploying agents in local or hosted environments
Pros
- Runs unprivileged agent loops without system access
- Supports local and client-server deployment modes
- Integrates MCP servers for tooling
- Allows customization via webhook workers
- Enables durable sessions with event logging
Cons
- No free tier beyond initial credits
- Requires configuration via TOML files
- Limited to TOML-based setup for agent definitions
- Self-hosting requires infrastructure management
Frequently asked questions about subs
What is subs and what does it do?
subs is a cloud-native agent harness designed to run unprivileged agent loops without system access. It supports local and client-server modes, integrates with MCP servers for tool access, and enables durable agent sessions with full event logging.
Who is subs designed for?
subs is designed for developers and teams who need to deploy AI agents in production environments, whether locally, self-hosted, or via a hosted service. It suits users who require flexible agent configuration and integration with external tools.
How do I get started with subs?
To get started, install the subs CLI, create a configuration file (e.g., subs.toml), and define your agent’s LLM provider, system prompt, and tools. You can then run the agent locally or deploy it using the hosted service.
Does subs support integrations with other tools?
Yes, subs supports MCP servers for tool integration, Slack for messaging, and webhook workers for customizing agent behavior. It also allows scoping MCP connections per user for personalized interactions.
Can I use subs with my own LLM?
Yes, subs allows you to bring your own LLM provider by configuring it in the subs.toml file. It supports various LLM providers through standardized interfaces.
What are the deployment options for subs?
subs can be deployed locally, self-hosted, or via the hosted service provided by substructure.ai. The hosted option allows you to run agents on their infrastructure while maintaining control through configuration files.
