Introduce Pluggable Context Engine

## What changed Hermes Agent added a pluggable context engine on May 18, 2026. Context handling now runs through modular hermes plugins instead of a fixed internal system. Users can swap engines to manage visibility rules, apply custom filters, control summarization, or inject domain-specific data.

The update extends the existing plugin system already used for memory providers and model switching. No core code changes are required to activate a new context backend.

## Why it matters Vibe Builders gain direct control over what information reaches the agent during long autonomous sessions. This reduces irrelevant noise from Telegram or Discord threads while allowing targeted pulls from Notion databases or Git logs.

The move pressures hosted agents that keep context logic closed and non-replaceable. It bets that self-hosted operators will invest time in custom context layers to improve reliability over months of use.

## How to use it Run the standard bash install for Hermes Agent on a VPS. Use the CLI command hermes plugin install to add a context engine package, then set the active engine in the YAML config or through the local web dashboard. Test changes by routing a multi-turn task and checking memory recall.

Plugins are available now through the official repository. No paid tier is required, though you still cover VPS and LLM token costs separately.

## Watch for Success shows up as measurably fewer context overflows in week-long workflows. The bet fails if plugin interfaces shift with every minor release. Expect a follow-up release that adds pluggable action filters next.