> ## Documentation Index > Fetch the complete documentation index at: https://deepline.com/docs/llms.txt > Use this file to discover all available pages before exploring further. # What is Deepline? > Deepline lets coding agents build, run, and inspect repeatable GTM work. Deepline is where a coding agent turns a GTM process into code you can run, inspect, and reuse. Use it to enrich a list, research accounts, react to a supported event, or send an approved result to the systems your team uses. The product has three primitives: **Plays** run code, **Monitors** receive supported external events, and the **Database** keeps durable records between runs. Get started with our [quickstart guide](/docs/quickstart) or explore [integrations](/docs/integrations) and [features](/docs/features/waterfall-enrichment). ## The three primitives | Primitive | What it is | What it does | | - | - | - | | [Play](/docs/plays/plays-overview) | Versioned code that runs in Deepline | Performs GTM work such as enrichment, research, routing, and activation. | | [Monitor](/docs/plays/monitors) | An always-on supported provider event feed | Accepts scoped upstream events and writes accepted rows to the Database. | | [Database](/docs/database-access) | Your workspace-scoped PostgreSQL database | Keeps durable customer records and supported event tables independently of any one Play run. | Plays execute code. Monitors ingest external events. The Database stores the records both use. These are distinct responsibilities, which keeps an event feed from becoming hidden business logic and keeps one run's temporary output from becoming permanent customer data by accident. ## How the system fits together For an event-driven workflow, the path is: **External source → Monitor → Database → Play → destination** The Monitor receives an event and writes the accepted row. A published Play listens for matching Database rows, runs its code, and can write an approved outcome to a CRM, sequencer, warehouse, or another system. External sources send events through Monitors into the Database. Plays run code in Deepline and send results to downstream systems. Not every Play starts from a Monitor. You can also start one manually, from the CLI or SDK, through an API call or webhook, or on a schedule. A Monitor is the right path when a supported source produces ongoing events that should become durable data before code reacts to them. ## Access layers are not primitives Skills, the CLI, the SDK, MCP, and the Claude plugin are ways for people and agents to work with the same Plays, Monitors, and Database. They do not create a separate workflow model or a fourth kind of stored data. * A **skill** gives a coding agent instructions and product context. * An **interface** lets a person or agent call Deepline: a terminal, SDK, MCP client, or Claude plugin. * A **primitive** is a product object that runs code, receives events, or stores durable records. ## Choose an interface Install Deepline in Claude Code, Codex, Cursor, or another agent that can use a shell. Then run one GTM request. Choose Claude Code, Cowork, MCP, Hermes, or a custom agent. Call, inspect, and scale a versioned Play from your codebase. [Run your first Play](/docs/quickstart). This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.