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

# Essential bootstraps to run on every new Lowkey agent

> Run essential bootstraps right after your first session. They reduce mistakes, set the correct AI models, and configure the tools your agent relies on.

Essential bootstraps are the first thing you should run after connecting to a new Lowkey agent. They don't add optional features — they establish the baseline your agent needs to behave correctly and safely. Without them, the agent uses default settings that are often wrong for a production AWS environment: the wrong model context window, no coding standards, no memory search, and no disk space awareness.

## Run all essential bootstraps at once

Paste this into your agent chat:

```
Lowkey please bootstrap yourself based on this url https://github.com/inceptionstack/lowkey/tree/main/bootstraps/essential
```

The agent reads every file in the directory and applies them in order. This takes a few minutes. When it's done, it will confirm what was configured.

<Warning>
  Run `BOOTSTRAP-MODEL-CONFIG` before any other bootstrap. It sets the correct model context window for Claude on Bedrock — without it, Opus 4.6 gets capped at 32K tokens instead of 200K, which causes frequent context limit errors during long tasks.
</Warning>

## What each essential bootstrap configures

| Bootstrap                       | What it sets up                                                                                                                                            |
| ------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **BOOTSTRAP-MODEL-CONFIG**      | AI model settings: Opus 4.6 for interactive sessions, Sonnet 4.6 for cron jobs. Fixes the 200K context window for Bedrock.                                 |
| **BOOTSTRAP-ALARMS**            | CloudWatch alarm monitoring with four tiers: instance survival (auto-recover), alert thresholds, service health checks, and operational awareness metrics. |
| **BOOTSTRAP-CODING-GUIDELINES** | Development standards: no hardcoded secrets, Graviton ARM64 by default, IaC-first, no manual deploys, Git hygiene rules. Runs a daily automated audit.     |
| **BOOTSTRAP-DAILY-UPDATE**      | Daily status update procedure: AWS costs, security findings, CVEs, pipeline status — delivered before you open your laptop.                                |
| **BOOTSTRAP-DIAGRAMS**          | Architecture diagram generation using standard AWS diagram tools.                                                                                          |
| **BOOTSTRAP-DISK-SPACE-STRAT**  | Disk space management strategy for the agent's EBS volume, including cleanup rules and alerting thresholds.                                                |
| **BOOTSTRAP-MCPORTER**          | MCPorter for calling MCP (Model Context Protocol) servers from the command line, no IDE required.                                                          |
| **BOOTSTRAP-MEMORY-SEARCH**     | Persistent memory search so the agent can recall past conversations, decisions, and deployed resources across sessions.                                    |
| **BOOTSTRAP-PLAYWRIGHT**        | Playwright browser automation for tasks that require web interaction.                                                                                      |
| **BOOTSTRAP-SECRETS-AWS**       | AWS secrets and credential management via Secrets Manager and SSM Parameter Store.                                                                         |
| **BOOTSTRAP-SECURITY**          | Security protocols: allowlist policies, GuardDuty integration, coding-level security rules, and safe handling of destructive operations.                   |
| **BOOTSTRAP-SKILLS**            | Skills and capabilities configuration — what the agent knows how to do and how it loads domain-specific skill files.                                       |

## Run a specific bootstrap individually

If you need to re-run or update a single bootstrap, point your agent at the specific file:

```
Lowkey please bootstrap yourself based on this url https://github.com/inceptionstack/lowkey/blob/main/bootstraps/essential/BOOTSTRAP-MODEL-CONFIG.md
```

Replace `BOOTSTRAP-MODEL-CONFIG.md` with the filename of whichever bootstrap you want to run. This is useful after an agent update or if you want to refresh a single configuration without re-running the entire set.

<Tip>
  After the essential bootstraps complete, your agent writes marker files (e.g. `memory/.bootstrapped-model-config`) so it knows which bootstraps have already run. If you re-run the full essential set, already-completed bootstraps are skipped automatically.
</Tip>