Instar: Add Persistent Memory and Telegram Chat to Your Claude Code Agent
What It Does
Instar is a persistent server that wraps Claude Code, giving it three critical capabilities it lacks out-of-the-box:
Persistent Memory & Identity: Your agent remembers conversations, decisions, and context across restarts. It maintains a coherent identity instead of starting fresh each session.
Scheduled Jobs & Autonomous Operation: Run Claude Code agents on schedules (health checks, email monitoring, reflection sessions) without manual intervention.
Messaging Integration: Talk to your Claude Code agent from Telegram or WhatsApp. Send it tasks, get responses, and have ongoing conversations from your phone.
This isn't an API wrapper—it runs actual Claude Code processes with full access to native tools, MCP servers, and sub-agents.
Setup: Running in Minutes
# 1. Run the setup wizard
npx instar
# 2. Start your agent
instar server start
# 3. Message it on Telegram
# (The wizard guides you through Telegram bot setup)
The wizard discovers your environment, configures messaging, sets up identity files, and gets everything running. You need Node.js 20+, the Claude Code CLI, and either an API key or Claude subscription.
How It Works Under the Hood
Instar solves what the creator calls "the six dimensions of agent coherence"—the gap between Claude Code's raw power and reliable autonomy. An agent that forgets yesterday's conversations or contradicts its own decisions can't be trusted with real tasks.
Your setup looks like this:
You (Telegram/WhatsApp/Terminal)
│
conversation
│
▼
┌─────────────────────────┐
│ Your AI Partner │
│ (Instar Server) │
└────────┬────────────────┘
│ manages its own infrastructure
│
├─ Claude Code session (job: health-check)
├─ Claude Code session (job: email-monitor)
├─ Claude Code session (interactive chat)
└─ Claude Code session (job: reflection)
Each session is a real Claude Code process with extended thinking capabilities. The agent manages all sessions autonomously.
Skills You Can Use Immediately
Instar ships with 12 skills following the Agent Skills open standard (portable across Claude Code, Cursor, VS Code, and 35+ other platforms). These are standalone—copy a SKILL.md into your project and they work with zero dependencies.
Instar-powered skills unlock capabilities needing persistent infrastructure:
- Scheduled code reviews that run automatically
- Project health monitoring with regular check-ins
- Learning sessions where the agent reflects on past decisions
- Task delegation with follow-up and accountability
Browse all skills at: agent-skills.md/authors/sagemindai
Security Model: Power User Infrastructure
Important: Instar runs Claude Code with --dangerously-skip-permissions. This is power-user infrastructure, not a sandbox.
Security lives in multiple layers:
- Behavioral hooks: Command guards block destructive operations before execution
- Safety gates: LLM-supervised review of external actions with adaptive trust
- Identity constraints: The agent's persistent identity includes safety boundaries
When To Use Instar vs. Other Tools
- OpenClaw: Excels at breadth—channels, voice, device apps, massive plugin ecosystem
- Instar: Focuses on depth—coherence, identity, memory, safety for long-running autonomous agents
Use Instar when you need Claude Code to be a reliable partner over time, not just a powerful one-off tool.
Getting Started Today
- Back up your current Claude Code projects (Instar will run new sessions)
- Run the setup:
npx instar - Test with a simple scheduled job: Set up a daily health check for your main project
- Try Telegram integration: Delegate small coding tasks from your phone
The full documentation at instar.sh includes installation guides, CLI commands, API endpoints, and configuration details.
