> ## Documentation Index
> Fetch the complete documentation index at: https://docs.sagescreen.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# AI automations: SageScreen agent fleet and CI/CD pipeline

> Topology and responsibilities of the SageScreen AI agent fleet that powers the CI/CD pipeline, dev tooling, and business automations.

# AI Automations

SageScreen operates a fleet of autonomous AI agents that power the CI/CD pipeline, development
tooling, and business automation layer. The diagram below shows the full fleet topology as of v4.

## Fleet Diagram

<Frame caption="SageScreen Agent Fleet v4 — CI/CD and automation architecture">
  <img src="https://sagescreen-artifacts.nyc3.cdn.digitaloceanspaces.com/diagrams/sagescreen_fleet_v4_final.svg" alt="SageScreen agent fleet diagram showing the CI/CD pipeline agents, automation agents, and their relationships to infrastructure, MCP tools, and external services" style={{ width: "100%", maxWidth: "1200px" }} />
</Frame>

> If the diagram is not visible, you can open it directly at the
> [Spaces CDN URL](https://sagescreen-artifacts.nyc3.cdn.digitaloceanspaces.com/diagrams/sagescreen_fleet_v4_final.svg).

## Architecture Overview

The SageScreen agent fleet is a set of Claude-powered coding and operations agents running inside
a mono-repo CI/CD loop. Each agent is scoped to a single concern — code review, testing, deploy
validation, Linear ticket automation, outreach, or document generation — and communicates through
structured MCP tool calls rather than free-form messages.

### CI/CD Agents

| Agent           | Trigger                        | Role                                                 |
| --------------- | ------------------------------ | ---------------------------------------------------- |
| **Code Review** | Pull request opened or updated | Reviews diffs for correctness, style, and security   |
| **QA**          | CI on every push               | Runs test suites and reports failures back to Linear |
| **PR Agent**    | Draft PR created               | Enforces branch/commit/title conventions             |
| **SRE Agent**   | Deployment webhook             | Validates post-deploy health checks                  |

### Automation Agents

| Agent                 | Trigger                            | Role                                                          |
| --------------------- | ---------------------------------- | ------------------------------------------------------------- |
| **Linear Ticket Bot** | Linear webhook or `/schedule` cron | Triages issues, assigns labels, files sub-tickets             |
| **Outreach Agent**    | n8n workflow                       | Enrolls contacts in Reply.io sequences and syncs to Pipedrive |
| **Docs Agent**        | Linear issue creation              | Generates or updates Mintlify documentation pages             |
| **Image Agent**       | Content request                    | Generates and uploads brand assets to Spaces CDN              |

### MCP Tool Layer

All agents share a common MCP server at `mcp.sagescreen.net` which exposes tools for:

* Linear (issues, labels, comments)
* Pipedrive (contacts, deals, activities)
* Reply.io (sequences, enrollment)
* DigitalOcean Spaces (artifact storage and CDN)
* Productlane (changelogs, customer feedback threads)
* WordPress (posts, pages, media)

### Transparency and Auditability

Every agent action produces a structured audit entry in the Linear issue timeline or a GitHub
commit signed by the bot identity. No agent merges to `dev` or `main` without a human approval
step. Secrets are never passed through agent prompts; they are resolved at runtime from environment
variables injected by the CI runner.

## Source Attribution

Fleet diagram designed by the SageScreen engineering team. SVG hosted on DigitalOcean Spaces under
the `sagescreen-artifacts` bucket (region: `nyc3`).