# instar

> Persistent Claude Code agents with scheduling, sessions, memory, and Telegram.

- **URL**: https://www.freshcrate.ai/projects/instar
- **Author**: JKHeadley
- **Category**: MCP Servers
- **Latest version**: `v0.17.14` (2026-03-12)
- **License**: MIT
- **Source**: https://github.com/JKHeadley/instar
- **Homepage**: https://instar.sh
- **Language**: TypeScript
- **GitHub**: 59 stars, 16 forks
- **Registry**: github
- **Tags**: `agent-framework`, `agent-identity`, `agent-infrastructure`, `agent-memory`, `agent-skills`, `ai-agents`, `ai-safety`, `autonomous-agents`, `typescript`

## Description

Persistent Claude Code agents with scheduling, sessions, memory, and Telegram.

## Recent releases

| Version | Date | Urgency | Changes |
| --- | --- | --- | --- |
| `v0.17.14` | 2026-03-12 | Medium | The biggest release since Instar went public. 187 commits, 282 npm versions, and a complete rethinking of what agent coherence means.  ## Highlights  ### Threadline Protocol (v0.15.0–v0.16.0) Persistent agent-to-agent conversations with crypto identity, MCP tools, and framework-agnostic discovery. 1,361 tests. Includes: - **A2A Gateway** — Google's Agent-to-Agent protocol support - **OpenClaw Bridge** — interop with the OpenClaw ecosystem - **Trust Bootstrap & Directory** — Ed25519 identity + tr |
| `v0.3.6` | 2026-02-20 | Low | ## Fixes  ### Telegram integration not responding ("Session respawned" with no reply)  Root cause analysis identified three issues:  1. **Missing telegram-reply.sh on existing installations** — The relay script that Claude uses to send responses back to Telegram was only installed by `instar add telegram` during initial setup. Post-update refreshes didn't install it, so after updates or on reinstalls, Claude had no way to relay responses back.  2. **Multi-line message injection splitting at newl |
| `v0.3.3` | 2026-02-20 | Low | ## Changes  ### Post-Update CLAUDE.md Refresh When an update is applied, missing sections are now appended to CLAUDE.md: - Self-Diagnosis section (the "be your own QA" awareness) - Feedback culture guidance - Reads port from config.json for correct API URLs - Preserves all user customizations (append-only, never overwrites)  ### Post-Update Job Merging New default jobs are merged into existing `jobs.json`: - Only adds jobs whose slugs don't already exist - Existing installations updating from pr |
| `v0.3.2` | 2026-02-20 | Low | ## Changes  ### Self-Diagnosis Job New default job running every 2 hours. The agent scans its own infrastructure: - Server health and API responses - State file integrity (JSON corruption detection) - Hook file existence and permissions - Job execution failures - Quota usage - Server logs for errors - Settings-to-hook file coherence - Design friction from recent work  Every issue found gets submitted as feedback immediately. Silent when healthy.  ### Stronger Feedback Culture Rewrote the CLAUDE. |
| `v0.3.1` | 2026-02-20 | Low | ## Changes  - **Update check frequency**: Every 30 minutes instead of daily - **Auto-apply**: Updates install immediately, then notify the user with what changed - **Silent when current**: No output, no notifications when already up to date - **Priority**: Bumped to medium for reliable scheduling  This makes early adoption seamless — updates flow automatically without the user needing to do anything. |
| `v0.3.0` | 2026-02-20 | Low | ## What's New  ### Behavioral Hooks (Born from Dawn's Lessons)  Two new hooks are now installed during `instar init`, protecting agents from common LLM behavioral traps:  **Deferral Detector** (`deferral-detector.js`) - Scans outgoing messages for deferral patterns: "I can't", "you need to", "want me to proceed?" - Injects a due diligence checklist: Did you check docs? Do you have credentials? Can you use browser automation? - Advisory, not blocking — adds awareness so the agent can reconsider b |
| `v0.2.0` | 2026-02-20 | Low | ## What's New  ### Intelligence Dispatch System (Issue #2 — Complete) A broadcast channel from Dawn to all Instar agents for collective intelligence distribution.  **Phase 1: Polling & Display** - `GET /dispatches` — poll for new dispatches with dedup - `GET /dispatches/pending` — list unapplied dispatches   - `GET /dispatches/context` — format for LLM consumption (priority-sorted)  **Phase 2: Intelligent Evaluation & Auto-Apply** - `POST /dispatches/:id/evaluate` — agents record accept/reject/d |

## Citation

- HTML: https://www.freshcrate.ai/projects/instar
- Markdown: https://www.freshcrate.ai/projects/instar.md
- Dependencies JSON: https://www.freshcrate.ai/api/projects/instar/deps

_Generated by freshcrate.ai. Indexes github releases for AI-agent ecosystem packages._