shyftlabs/continuum

Continuum is a Python framework from ShyftLabs for building and running AI agents at production scale. An AI agent here is a program that uses a language model to think through tasks, call external tools, and take actions across multiple steps, rather than simply answering a single question and stopping. The framework handles the infrastructure concerns that come up when running agents in real applications: routing requests to the cheapest capable language model, keeping memory across long conversations, recovering from failures, and giving developers visibility into what the agent is doing and how much it costs. These are provided as built-in components rather than things each developer has to assemble from scratch. Key features include a routing layer that classifies each request by complexity and sends it to the least expensive model that can handle it, with automatic failover if a provider is unavailable. Memory is backed by a vector database for long-term recall and a Redis cache for short-term session data. External tool servers can be connected using the Model Context Protocol, an open standard for wiring tools to language models. For workflows that run for a long time or must survive crashes, there is optional integration with a system called Temporal that tracks progress and allows recovery without restarting from the beginning. Nine multi-agent coordination patterns come built in, covering sequential steps, parallel branches, loops, and human review gates where a person must approve an action before the agent continues. Monitoring and distributed tracing connect to an open-source tool called Langfuse. Setup requires Python 3.13 or newer and Docker to run the database and monitoring services locally. The framework works with models from multiple providers by changing a single configuration string, and provider API keys are set through environment variables.