Your agents work
the night shift
Self-hosted AI agent fleet. Create agents, give them skills, plug them into your channels. You sleep, they ship.
Quick start
Run the fleet in one command
Start with npx. Move to source or Docker when you need deeper control.
Downloads runtime, runs migrations, and starts the daemon.
$npx @nitejar/cli upWhat's inside
One place to run the whole fleet
Everything you need to create, connect, observe, and manage your AI agents.
Agents that work together without stepping on each other
Put multiple agents in the same channel. Each one triages incoming messages independently — responds when relevant, stays silent when not. No routing rules, no orchestration config. The first agent to claim a message gets exclusive access. Check the activity timeline to see every triage decision.
semantic triage · silent pass · exclusive claims · activity log
Command Center
Live fleet health, active operations, queue depth, and cost posture at a glance.
Stat cards · agent roster · sparkline trends
Activity
Cross-channel run history. Every execution, every step, searchable and filterable.
Run timeline · status tracking · source badges
Agents
Full agent builder. Identity, personality, model, skills, plugins, network policy, budgets — all in one config.
Soul config · tool access · budget limits
Collections
Shared structured data with schema governance. Give agents access to the knowledge they need.
Schema enforcement · per-agent access control
Costs
Spend trends, per-agent breakdowns, source attribution. Set soft and hard budget limits.
Daily trends · org/team/agent scopes
Plugins
Connect agents to GitHub, Telegram, webhooks, and more. Install community plugins or build your own.
Channel integrations · custom plugins
Skills
Reusable knowledge packs, workflow templates, and scripts. Synced to each agent's sandbox.
Knowledge · workflows · scripts
Evals
Scoring pipelines and quality trends. Measure how well your agents are actually performing.
Evaluators · score history · quality trends
Sessions
Talk to one or more agents directly in-app. Real conversations, not just config and deploy.
Multi-agent chat · in-app conversations
Use cases
What your fleet does while you're away
Triage, draft, research, summarize, sync, alert. You name the workflow — your agents run it.
How it works
Three steps to the night shift
Set up in minutes. The complexity lives in what your agents do, not in getting them running.
Create agents
Give each agent a name, a personality, skills, and a budget. Full config — identity, model, tools, network policy — all in one place.
Connect channels
Plug agents into Telegram, GitHub, webhooks, or talk to them directly in-app. One agent, many surfaces.
3 repos · issues + PRs
2 groups · DMs enabled
1 endpoint configured
Watch the fleet
Live activity, cost trends, agent-by-agent breakdowns. See what each agent did, what it spent, and what it's doing next.
Posted diagnosis on issue #347
Sent weekly digest to #engineering
Opened PR #412 — fix auth token refresh
Reviewing PR #409
Multi-agent
One channel. Multiple agents. Zero crosstalk.
Agents triage messages independently, claim exclusive turns, and stay silent when it's not their domain. No routing rules — just distinct roles and semantic understanding.
Getting 502s on the checkout API. Customers are seeing payment failures. Can someone look?
On it. Seeing elevated error rates starting 2:08 PM. The payment-service pod is restarting in a loop — OOM kill. Last deploy bumped the batch size from 100 to 5000. Rolling back now.
Confirming from monitoring: p99 latency spiked 12x at 2:08. Memory usage hit 98% on payment-service-7b. Three other services are healthy. I'll set an alert if error rates don't drop after the rollback.
Command center
See everything your agents do.
Live fleet health, active operations, cost tracking. One screen to know exactly what's happening.
Agents
3
Runs today
47
Success
96.8%
Spend
$4.20
| Agent | Status | Runs | Cost |
🦉nightowl | busy | 142 | $1.23 |
🔧fixer | idle | 89 | $0.87 |
📡scout | busy | 231 | $2.10 |
Triaging issue #347
githubWeekly digest draft
schedulerOpen source
Built in the open
Self-hosted, Apache-2.0, no vendor lock-in. Run it on your infrastructure, fork it for your needs.
GitHub
Star the repo, read the source, open a PR. Apache-2.0 — fork it, ship it, own it.
View repositoryDiscord
Join the conversation. Share agent setups, debug workflows together, show what your fleet built overnight.
Join DiscordDocs
Setup guide, agent configuration, plugin API. Everything you need to run the night shift.
Read the docs