Practical overview
When it fits
Use it when an agent workflow spans sessions and needs archived history, layered recall and curated knowledge notes without modifying the agent core. Its evaluation and governance features also suit operators who need to inspect registry quality or edge plans.
Who it is for
- Operators of local AI-agent setups who want session archiving and persistent recall.
- Technical users running Hermes, Claude Code, Codex-style agents or Cursor-style shared data-directory setups.
- Teams reviewing memory quality, registry policies or proposed gbrain edge plans.
Documented uses
- Install a memory sidecar into an agent's data directory.
- Archive sessions and organize durable knowledge for later retrieval.
- Supply recall from hot, warm, cold and curated knowledge layers.
- Review memory-quality trends, lint private registries and inspect policy validity or conflicts.
- Preview gbrain edge plans before applying them.
Requirements & limits
The project is a public GitHub repository with an MIT license. The documented installation interface is a CLI.
- The required infrastructure is specific: Python 3.9+, PostgreSQL 16, Hindsight, gbrain and agent state/session files.
- Installation targets an agent directory through AGENT_HOME and deploys scripts beneath that directory.
- The sidecar uses hot, warm and cold recall layers alongside curated knowledge notes.
- The installer offers automatic, guided and detection-only dependency-assistance modes.
- The memory-quality cron block starts paused and requires operational review before unpausing.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Install the sidecar into an agent data directory
Use the documented installer flow to target an agent directory and deploy the sidecar.
Before you start
- Python 3.9 or later
- PostgreSQL 16
- Reachable Hindsight and gbrain services
- An agent data directory containing state.db and session files
Steps
Clone the repository and change into its directory.
Set AGENT_HOME to the agent data directory, such as $HOME/.hermes.
Run ./install.sh.
Expected outcome
The installer deploys the public sidecar runtime and related assets for the selected agent directory.
Guide limitations
- The documented path assumes the required services and agent data directory are available.
- The two Hermes-oriented maintenance helpers memory_watermark.py and memory_snapshot_backup.py are not installed by default in the public multi-agent path.
Run the documented post-install maintenance checks
Resume session processing, run a maintenance cycle and invoke the sidecar acceptance check after installation.
Before you start
- A completed installation with AGENT_HOME set
- Python 3 and the installed scripts under $AGENT_HOME/scripts
Steps
Run python3 "$AGENT_HOME/scripts/session_to_gbrain.py" --resume.
Run python3 "$AGENT_HOME/scripts/memory_maintenance_cycle.py".
Run python3 "$AGENT_HOME/scripts/sidecar_acceptance_check.py".
Expected outcome
The documented post-install sequence resumes session processing, runs maintenance and invokes the acceptance-check script.
Guide limitations
- The README provides the commands but does not specify expected output or acceptance criteria.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 14 October 2026
- Original website title
- mage0535/hermes-memory-installer · Official homepage title · retained in its source language