Instar is a server-based platform designed to support the creation of persistent, self-evolving AI agents with long-term memory, continuity, and autonomous operation. The system is built to address common shortcomings in agent frameworks, such as forgetting past interactions, losing context, breaking after updates, and lacking consistent identity across communication channels. Instar agents are architected to remember previous conversations, track relationships, maintain commitments, and evolve their own capabilities over time.
A key feature of Instar is its structural enforcement of six core conditions for trustworthy autonomy: memory retention across sessions using SQLite and rolling summaries, cross-platform identity resolution, continuity of agent identity through restarts and updates, temporal awareness via event tracking, consistency in commitments through decision journaling and coherence gates, and an evolution system for ongoing learning and self-improvement. The platform integrates with Claude Code or Codex subscriptions, running as a server on the user's machine to manage Claude Code sessions for various tasks. It supports two-way messaging via Telegram, mapping each forum topic to a session, and provides a dashboard-like experience within group chats.
Instar includes a job scheduler that allows users to define tasks in JSON, set cron-based schedules, assign priorities, and manage concurrency and quotas. md for learned information, and configuration files for server and scheduled jobs. Safety mechanisms are built in, such as LLM-supervised operation gates, decision journaling, drift detection, and adaptive trust per service, ensuring that each decision is recorded and that dangerous commands are guarded against.
Designed for users who require reliable, autonomous AI agents capable of continuous operation and self-improvement, Instar provides the foundational infrastructure extracted from a production AI system. It is positioned as coherence infrastructure for self-evolving agents, enabling developers to build agents that can operate with autonomy and resilience across various platforms and communication channels.
Instar sits in PulseGate's AI & ML category. Enabling developers to build AI agents that retain memory and evolve autonomously over time. It is built as a B2B product for AI developers and advanced users building autonomous agents. Instar is a paid product. It runs on the command line.
Instar first shipped in 2026. Development happens publicly on GitHub with 68 stars and 2.8k commits in the last 90 days. Key capabilities include persistent memory, autonomous agent, and Claude Code integration.
Latest indexed changes and source events
Other apps tracked under the same category.