Releases·Tool Release
Eidra Agent ships v0.1 CLI prototype for persona agents with persistent memory and adversarial evaluation
Vigorlee released Eidra Agent v0.1, an MIT-licensed Node.js CLI prototype for persona agents combining configurable personas, explicit persistent memory, source-tagged lexical retrieval, adversarial checks, and preference data export. It targets the research question of preserving agent identity and traceable behavior, positioning failures as reviewable improvement data for post-training.
27 engagement1 sourceWed, Sep 9, 2026, 05:21 AM
AI Overview
- GitHub repo vigorlee/eidra-agent released v0.1, a Node.js 22+ CLI prototype with MIT license and CI, under 27-engagement post.
- v0.1 implements JSON-defined personas, lexical retrieval with source tags, and explicit memory that persists across restarts and supports inspection.
- Adversarial smoke tests and preference-data export turn interaction failures into reviewable training data.
- Autonomous planning, model-directed tool use, multi-agent games, training, and multimodal interaction remain roadmap items.
- The default mock endpoint returns fixed responses to verify workflow, not model capability or robustness.
Entities
Related
Developments·Paper
Emergi-PersonaOS: A Persona Agent OS for Situational Adaptation and Controllable Evolution
An arXiv paper introduces Emergi-PersonaOS, an operating system for managing persona agents across their lifecycle. It uses a three-layer persona representation to handle situational adaptation and controlled long-term evolution, demonstrated on television-character dialogue. This gives practitioners a framework for agents that maintain identity continuity while adapting to context.
An arXiv paper introduces Emergi-PersonaOS, an operating system for managing persona agents across their lifecycle. It uses a three-layer persona representation to handle situational adaptation and controlled long-term evolution, demonstrated on television-character dialogue. This gives practitioners a framework for agents that maintain identity continuity while adapting to context.
0 engagement1 sourcearxivWed, Sep 23, 2026, 06:34 AM
Social·Community
Two-person team publishes disclosure of AI agent harness protocol for identity-constitutive memory and alignment
An author and colleague published a technical disclosure on TDCommons describing an AI agent harness protocol intended to give agents persistent, evolving identity traits and 'identity constitutive memory' while reducing sycophantic drift and confabulation. The team claims consistent results across multiple AI platforms and says they use it daily. The disclosure is an enabling publication only; the working documents are held as trade secrets with possible commercial and non-commercial licensing later. The claims are self-reported and unverified by third parties.
An author and colleague published a technical disclosure on TDCommons describing an AI agent harness protocol intended to give agents persistent, evolving identity traits and 'identity constitutive memory' while reducing sycophantic drift and confabulation. The team claims consistent results across multiple AI platforms and says they use it daily. The disclosure is an enabling publication only; the working documents are held as trade secrets with possible commercial and non-commercial licensing later. The claims are self-reported and unverified by third parties.
4 engagement1 sourceredditTue, Oct 6, 2026, 07:42 PM
Products·launched
Persistent AI agent with memory, skills, and Obsidian archive for ongoing work
A persistent AI agent built on Hermes Agent that maintains identity, selective memory, and behavioral rules across sessions. It uses an Obsidian archive for durable work storage, integrates with tools like Linear and Xero, and performs scheduled tasks such as morning briefs and backups, solving the problem of having to re-explain context and preferences in every new chat.
A persistent AI agent built on Hermes Agent that maintains identity, selective memory, and behavioral rules across sessions. It uses an Obsidian archive for durable work storage, integrates with tools like Linear and Xero, and performs scheduled tasks such as morning briefs and backups, solving the problem of having to re-explain context and preferences in every new chat.
14 engagement1 sourceredditSun, Jul 19, 2026, 11:57 AM
Products·building
Argus: agentic runtime for long-term autonomous research agents with persistent memory and verification
Argus is a general agentic runtime built on top of the Harness architecture for autonomous research systems. It uses LLM agents that aim for continuous learning and self-evolution over long-running tasks without model weight updates, relying on persistent memory, external verification, and accumulated experience. It targets researchers exploring long-term autonomous agent operation.
Argus is a general agentic runtime built on top of the Harness architecture for autonomous research systems. It uses LLM agents that aim for continuous learning and self-evolution over long-running tasks without model weight updates, relying on persistent memory, external verification, and accumulated experience. It targets researchers exploring long-term autonomous agent operation.
1 engagement1 sourceredditWed, Sep 2, 2026, 10:16 AM
Releases·Tool Release
Hermespace: persistent agent world for Hermes agents released on GitHub
Hermespace is an append-only persistent world for Hermes agents, recording every session, belief, landmark, and evolution without pruning or decay. It provides a pocket workbench with limited FOA desk, dual decode, skills+MEMORY fabric, and neural FOA, enabling agents to build a deepening model of themselves and their environment across sessions.
Hermespace is an append-only persistent world for Hermes agents, recording every session, belief, landmark, and evolution without pruning or decay. It provides a pocket workbench with limited FOA desk, dual decode, skills+MEMORY fabric, and neural FOA, enabling agents to build a deepening model of themselves and their environment across sessions.
69 engagement1 sourcegithubSat, Jul 18, 2026, 01:23 PM