- Home
- Skills
- Agents & Orchestration
- easy agent mem - Persistent Coding Memory
More screenshots
Works with the AI Agents you already use
easy agent mem - Persistent Coding Memory
- Hand off to a fresh chat: save outcomes, changed files, blockers and next steps, together with the durable session handle.
Free
easy agent mem - Persistent Coding Memory
Example session with this skill installed
In this demo project, remember our accepted decision to use SQLite for the single-user prototype because it must run without a database service. Use docs/architecture.md as evidence. Save a handoff, reconnect to the same private session and recall the decision. Then edit the evidence file to put the storage choice under review, check its freshness and confirm the original decision is preserved. Continue saving and updating throughout the project.
- Read your context and instructions
- Compiled the easy agent mem
Saved and recalled the storage decision
Recorded local MCP demonstration using easy-agent-mem 0.8.0 and a synthetic repository. This report was assembled from captured tool responses, rather than a recorded autonomous-agent chat.
Saved decision
Use SQLite for the single-user prototype because it must run without a database service.
| Field | Recorded value |
|---|---|
| Kind | decision |
| Status | accepted |
| Source | Synthetic demo architecture decision in docs/architecture.md |
| Evidence | docs/architecture.md, captured with a content hash |
| Scope | Private session |
| Record ID | 763b46ede53aa7de5f3fb0ba7d8edcf10bf8b31a87aee07121c3788fc60fb69c |
Observed results
| Operation | Result |
|---|---|
| Save the decision | Stored successfully through record_decision |
| Save the handoff summary | Saved to local scoped engineering memory |
| Restart the MCP process | New connection initialized successfully |
| Recall with the original session handle | The saved SQLite decision was returned |
| Inspect evidence before editing | Freshness was current |
| Edit the evidence file | Storage selection was put under review for multi-user deployment |
| Check the remembered decision | Freshness became needs_review |
| Inspect historical wording | The original SQLite decision remained unchanged |
| Query a different private session | The private decision was not returned |
Continue the private session
{
"project_name": "demo-project",
"query": "Resume prototype storage implementation",
"session_id": "agensi-demo-20261104"
}
Reuse these arguments in the same repository, worktree and branch.
Next steps
Read the changed architecture document and decide whether the multi-user requirement replaces the original SQLite choice. Record a replacement only after that decision is established.
The freshness flag alone does not establish a contradiction. No application implementation or application tests were claimed in this demonstration.
Connects securely to your tools. The creator never sees your data.
About this skill
Pick up where your last coding session stopped
A fresh chat should be able to find why you chose a database, what still blocks the task, and which file supports a decision.
This skill teaches your coding agent to use easy-agent-mem for focused project recall, sourced decisions and structured handoffs. The free, separately installed local MCP server provides the storage.
What it does
- Resume project work: retrieve relevant saved context before continuing a task.
- Remember decisions: capture agreed decisions, constraints and blockers with their source and supporting files.
- Check evidence: flag when a supporting file has changed or disappeared so the decision can be reviewed.
- Hand off to a fresh chat: save outcomes, changed files, blockers and next steps, together with the durable session handle.
- Keep an optional Obsidian record: export to a configured vault, or use local storage without Obsidian.
How it works
Connect your agent to Agensi and to your local agent-mem MCP server. Ask it to recall, remember or hand off project work. The skill guides the relevant memory calls and returns useful context with record IDs and evidence paths.
Private continuity requires the same repository, worktree, branch and session handle. Sharing a record more widely is explicit.
Included
One SKILL.md workflow, setup and troubleshooting guidance, and concrete tool-call examples. Install the CLI/MCP runtime separately from PyPI.
Requirements
Python 3.10+, easy-agent-mem 0.8.0 or later exposing the required tools, and a connected local agent-mem MCP server. Obsidian is optional.
Basic local memory operations need no separate LLM API key. Your normal agent/model usage costs still apply.
Scope
The skill operates when invoked. It does not install lifecycle hooks or guarantee automatic saving at exit.
Changed evidence means review is needed; it does not prove a decision is wrong. Saved context still needs to be checked against current instructions and code.
The included demonstration exercised save, reconnect, recall and changed-evidence behavior through real MCP calls against PyPI 0.8.0 in a synthetic local repository. Individual IDE conversations were not evaluated for this package.
Requires a separately installed and connected local agent-mem MCP server. This ZIP does not install the runtime, native hooks or IDE configuration.
Private records require the original repository, worktree, branch and session handle. Retrieval is fallible; changed evidence signals a need for review rather than proving a contradiction.
The recorded demonstration used MCP stdio on macOS with a synthetic repository. Live IDE conversations, Windows/Linux and configured Obsidian export were not exercised for this package. Automatic saving at exit and measured token savings are not guaranteed.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 13 days ago
- Free with an Agensi account
Requires
Permissions
Allowed Hosts
File Scopes
The local agent-mem MCP server reads task-relevant project evidence and writes memory state under .agent-memory/** in the selected repository. With Obsidian configured, it also exports under Memory/Agent-Mem/** inside the selected vault. Evidence paths depend on the task; docs/** and src/** are common examples. Terminal and network access support installation and diagnostics. Core local save/recall does not require a separate remote model API. The skill does not need raw chat archives, .env files or credentials. Your coding agent/model provider’s normal processing still applies.