activegraph

ActiveGraph shipped seven runtime releases (1.2.0 to 1.7.1) to harden its reference packs for autonomous agents, adding a safe fork-test-promote loop with subprocess isolation, manifest hashing, and complete event logging.

ActiveGraph is an event sourcing runtime that treats the append-only event log as the agent itself. It achieves deterministic replay, forking experiments, and end-to-end traceability through graph projection, providing a new architecture for long-running, auditable agent systems.

Yohei Nakajima is presenting about Activegraph at AI House.

ActiveGraph announces two new papers on agent memory (LongMemEval) and self-improvement regimes, along with reference agents, pack templates, and upcoming meetups in Seattle and San Francisco.

The author showcases a controlled self-improvement approach for AI agents using a regime-to-seam method where failures are categorized to fix targeted areas, built on activegraph.

This paper introduces Regimes, an auditable, held-out-gated improvement loop built on the ActiveGraph runtime for self-improving agents. It demonstrates modest improvements on the LongMemEval dataset by autonomously discovering prompt repairs that pass static checks, sandbox execution, and held-out validation.

Regimes is an auditable, held-out-gated improvement loop built on the ActiveGraph event-sourced runtime. It diagnoses failures in AI agents, proposes repairs, and promotes them only after passing multiple gates, improving accuracy on LongMemEval by up to +0.10.

Yohei Nakajima releases an interactive site teaching ActiveGraph concepts and AG Coder, an open-source reference coding agent built on ActiveGraph.

Yohei Nakajima shares a demo of a coding agent on activegraph that automatically produces a trace and graph.

A deep research agent based on active graphs that provides a full graph trace of claims, sources, and agent activity.

Yohei Nakajima highlights that building an agent on activegraph automatically produces first-class traces, unlike bolted-on solutions, demonstrated with a coding agent experiment.

In our second longmemeval experiment, we introduce semantic ingestion into recall leveraging the ActiveGraph runtime, improving retrieval from 60.6% to 83.4%/84.8% for flat/agentic retrieval with LLM ingestion.

ActiveGraph introduces a deterministic non-generative approach for evidence compilation before semantic memory, achieving 85.6% QA accuracy and 86.2% turn answer-in-context on LongMemEval-S.

Yohei Nakajima celebrates the first citation of ActiveGraph on arXiv, where it is referenced as a complementary runtime, and notes that a working bridge example has been included in the repo.

Yohei Nakajima ran the LongMemEval benchmark on ActiveGraph, achieving 85.6% QA accuracy and 86.2% turn answer-in-context, demonstrating the effectiveness of event-based agent systems for long-term memory.

A developer built a fully traceable and forkable research agent using Active Graph and monid_ai, ensuring every claim is natively traced to its source, and got it working in about 30 minutes.