Introduction

Nitejar is a self-hosted AI agent fleet for teams that want automation with clear evidence. Your agents can triage, draft, sync, and alert across Telegram, GitHub, Slack, Discord, and webhooks. Every run leaves a traceable activity timeline, execution log, and cost breakdown.

Features

Nitejar focuses on operational workflows, not toy demos. Core capabilities today:

• Fleet operations: Configure multiple agents with separate identity, soul, tools, and limits.
• Channel integrations: Receive and respond through Telegram, GitHub, Slack, Discord, and generic webhooks.
• Skills: Attach reusable playbooks so agents follow your team’s process.
• Tools and sandboxes: Give agents controlled execution and data access.
• Collections and memory: Persist structured data and long-lived context.
• Evals: Score quality and catch behavior regressions.
• Routines: Schedule recurring runs for daily or weekly workflows.
• Credentials: Store secrets safely and scope access per agent.
• Costs and budgets: Track spend per run, model, and agent with budget controls.
• Receipts and debugging: Audit what happened, when, and why from the timeline.

Start Here

If this is your first run, start here:

• Quickstart
• Configuration
• Your First Agent

Operate Nitejar

For day-to-day operator workflows:

• Use Nitejar
• How Nitejar Works
• Integrations
• Costs and Budgets
• Receipts and Debugging

Build on Nitejar

For plugin authors and contributors:

• Build on Nitejar
• Architecture and Runtime
• Plugin SDK
• Skills Authoring
• Contributing