Your Coding Agent Can Read the Code—but Can It See the App Fail?

TailFlow is an open-source tool designed to help coding agents detect and diagnose issues with applications even after they have been edited. While coding agents are becoming capable of navigating repositories, editing multiple files, running tests, and explaining unfamiliar systems, they still face a gap in being able to observe what happens after an application starts.

TailFlow addresses this issue by providing a local runtime-verification layer for coding agents. It collects output from various sources such as development processes, local Docker containers, log files, and piped standard input. This runtime view is then exposed through different interfaces like MCP tools for coding agents, a shell CLI, an interactive terminal UI, a local web dashboard, and an HTTP API with Server-Sent Events stream.

The main goal of TailFlow is not just to display logs, but to let an agent answer specific questions about the application's behavior after an edit. It determines if expected services started, if a change triggered a successful rebuild, what failed after a specific change, and if the service is healthy or never started. It helps identify if an error is new or old, if the service is healthy, if a single error is repeated, or if the issue is related to hot reload or background jobs.

Tests are still essential, but TailFlow goes beyond them by providing evidence that static analysis and isolated tests cannot offer. It lets the agent access the application's runtime output, which can reveal issues that tests can't catch. For example, it can confirm if environment variables are available, if ports can be bound, if a container can communicate with other services, if a development server completed startup, if hot reload succeeded, and if a background job finished.

Getting started with TailFlow is straightforward. You can install it through npm by running 'npm install -g tailflow'. This installs four commands: 'tailflow' for an interactive TUI and project initializer, 'tailflow-daemon' as a runtime collector and local API, 'tailflow-mcp' as an MCP bridge for coding agents, and 'tailflow-logs' as a shell client for queries and automation.

After installation, you can initialize TailFlow in your project root with 'tailflow init'. It detects common runtime sources like 'dev', 'serve', and 'start' scripts in 'package.json', Pnpm, Yarn, Bun, and npm project metadata, Docker Compose files, and common local log directories. It then proposes a configuration and writes a 'tailflow.toml' file. Existing configurations are never replaced unless '--force' is explicitly provided.