@estebanforge/pi-agentmemory
Pi-native agentmemory tools — cross-session memory via REST, no MCP dependency
Package details
Install @estebanforge/pi-agentmemory from npm and Pi will load the resources declared by the package manifest.
$ pi install npm:@estebanforge/pi-agentmemory- Package
@estebanforge/pi-agentmemory- Version
1.0.0- Published
- Jun 10, 2026
- Downloads
- 91/mo · 91/wk
- Author
- estebanforge
- License
- MIT
- Types
- extension
- Size
- 15 KB
- Dependencies
- 0 dependencies · 2 peers
Pi manifest JSON
{
"extensions": [
"./extensions/agentmemory"
]
}Security note
Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.
README
@estebanforge/pi-agentmemory
Pi-native agentmemory tools. Bypasses MCP entirely — direct REST to the agentmemory server with automatic recall injection.
Install
pi install npm:@estebanforge/pi-agentmemory
Or install directly from git:
pi install git:https://github.com/EstebanForge/pi-agentmemory
Requires the agentmemory server running locally. See agentmemory for setup.
Tools
| Tool | Description |
|---|---|
memory_health |
Check whether the agentmemory server is reachable and healthy |
memory_search |
Search memory for prior decisions, preferences, bugs, and workflows |
memory_save |
Save a durable fact, convention, or bug fix into memory |
Hooks
| Hook | Behavior |
|---|---|
session_start |
Derives session ID, checks server health, sets footer status |
before_agent_start |
Auto-recalls relevant memories via smart-search, injects into system prompt |
agent_end |
Observes conversation turn back to agentmemory for future recall |
Command
| Command | Description |
|---|---|
/agentmemory-status |
Check server health from the command palette |
Configuration
Set environment variables (preferred):
export AGENTMEMORY_URL=http://localhost:3111 # default
export AGENTMEMORY_SECRET=your-secret # optional
export AGENTMEMORY_REQUIRE_HTTPS=1 # enforce HTTPS for non-loopback
How it works
On every prompt, the extension searches agentmemory for relevant context and injects it into the system prompt. After the agent responds, the conversation turn is observed back to agentmemory so future sessions can recall it.
No MCP server required. The extension talks directly to the agentmemory REST API.