Enso
amazinglvxw/enso-os
Discipline plugin with code-enforced learning, active forgetting and proactive challenges for Hermes
Enso is a discipline plugin for AI agents that turns errors into lessons enforced by hooks, prunes stale knowledge and adds a proactive challenge at the start of a session. It installs into Hermes Agent, Claude Code, OpenClaw and Gemini CLI with a single script flag.
What Enso does
Enso describes itself as an AI plugin that speaks first. It adds three things: code-enforced learning, where errors become hooks rather than prompts; active forgetting, where stale knowledge is pruned; and PAC, the Proactive Accountability Challenge, added in v0.7.0. In practice, an error in session one is captured automatically, one to three lessons are distilled when the session ends, and those lessons are injected into session two so the agent avoids the same mistake.
PAC watches session logs, memory files and decision patterns and, when it detects self-limiting behavior the user has not asked about, writes a Socratic challenge delivered at the start of the next session. It looks for five patterns: repetition, claim-action conflict, capability-task mismatch, sunk cost and critical decision nodes. It separates constraint-optimal choices, which it affirms, from self-limiting ones, which it challenges, and defaults to silence when unsure. Installation is a bash script that needs bash and python3, with targets for Claude Code, Gemini CLI, Hermes, OpenClaw and any agent with lifecycle hooks, and the README says data stays local.
Key features
- Errors captured automatically and distilled into one to three lessons per session
- Lessons injected into later sessions through hooks
- Active forgetting that prunes stale knowledge
- PAC challenges based on five detected behavior patterns
- Install targets for Hermes, Claude Code, Gemini CLI, OpenClaw and generic hook-capable agents
- Local data and a shell-script install
When to use it
- Stopping an agent from repeating the same error across sessions
- Getting a challenge at session start when logs show a pattern of abandoned projects
- Adding a hook-based learning layer to Hermes using only bash and python3
Who it is for: Hermes Agent users who want lessons from past errors enforced automatically and an AI that raises blind spots without being asked.
How it fits with Hermes Agent
Supports Hermes Agent as one of several install targets, through bash install.sh --target hermes.
How to install Enso
These commands are copied from the project's README. Check the repository for the latest steps before you run them.
git clone https://github.com/amazinglvxw/enso-os.git
cd enso-os
bash install.sh --target hermesRequirements: bash and python3 (pre-installed on macOS and Linux)
FAQ
What is Enso?
Enso is a discipline plugin for AI agents. It enforces lessons from past errors through hooks, prunes stale knowledge, and uses PAC to raise self-limiting patterns at the start of the next session.
How do I install Enso for Hermes Agent?
Clone the repository, change into the enso-os folder and run bash install.sh --target hermes. Then start a new session and Enso is active.
Is Enso free and open source?
Yes. The repository is released under the MIT license.
Similar plugins for Hermes Agent
All pluginsCompression proxy that replaces bulky tool output with short previews before it reaches Hermes Agent
BkashJEE Hermes Bot ForgeHermes plugin that builds a complete Bot from one sentence, with SOUL.md, memory, tools and a gateway
using-system otelysseyMarketplace of OpenTelemetry plugins for coding agents, installable in Hermes with a plugin pack
mykeura Minimalist Themes for Hermes DesktopEighteen warm, light color themes for Hermes Desktop, installable as a Hermes plugin
rusty4444 Hermes Home Assistant Voice IntegrationHome Assistant custom integration and Hermes plugins for smart-home control and a local voice loop
rarf Hermes QuotaPer-provider quota, rate limits and reset times in Hermes Desktop and the CLI, read from a local cache
Related guides: How to install Hermes Agent · How to run Hermes Agent securely