My Wiki!

Page locked

This page is currently locked for editing by another user. You have to wait until this user finishes editing or the lock expires.

Currently locked by:
Lock expires at: 2026/08/06 00:42 (15 min)

You've loaded an old revision of the document! If you save it, you will create a new version with this data.

# Kilo Code for VS Code (Insider)

Requirements:

* I want to have research agent (web) that is available for all project, sothat I can do structured web search on topics.
* The research agent should store it's research memory as text (.md) files in local project folder structure for to create growing context for the project.
## Custom Modes and Agents

Kilo discovers these files via glob patterns in multiple config directories:

Scope	Mode file path pattern      Agent file path pattern

Project-local	.kilo/modes/*.md         .kilo/agents/*.md

Global (XDG)	~/.config/kilo/modes/*.md       ~/.config/agents/modes/*.md

Legacy global	~/.kilo/modes/*.md or ~/.kilocode/modes/*.md
## Agent Memory
  * https://developers.openai.com/cookbook/examples/agents_sdk/building_reliable_agents_memory_compaction
  * https://voxos.ai/blog/how-to-give-ai-coding-agents-long-term-m/index.html
  * https://github.com/Kilo-Org/kilocode/discussions/2022
  * https://github.com/Kilo-Org/kilocode/discussions/377

Folder and Manifest Best Practices

Put source documents, manifests, helper files, and output directories in the Manifest instead of pasting large content into the prompt.

Put longer task instructions in workspace files such as README.md, task.md, or AGENTS.md; keep agent instructions focused on behavior and boundaries.

Use stable document IDs and a machine-readable manifest file so generated memos can cite sources and reviewers can inspect the path back to evidence.

Let Memory() manage its own memory artifacts. By default, sandbox memory uses memories/ and sessions/ under the workspace.

Keep generated artifacts under outputs/ so the application can inspect, copy, validate, or archive them after the run.

Keep mount scopes narrow. If you mount a data room, mount only what the agent should read or write.

Treat secrets as runtime configuration injected by your application or sandbox provider, not as prompt text or committed manifest content.

Prefer a small synthetic File(...) or Dir(...) entry for a tutorial, then switch to LocalDir, GitRepo, or storage mounts for production-sized datasets
.
### This is a visual preview of the sandbox workspace structure
.
### The next cell builds the actual Manifest entries manually.
```
WORKSPACE_TREE = """
/workspace/
  README.md
  manifest.csv
  docs/
    batch_1/
    batch_2/
    batch_3/
  outputs/
  memories/    # Generated by Memory()
  sessions/    # Generated by Memory()
""".strip()

print(WORKSPACE_TREE)

```

Prepare a Small Evidence Workspace
### memvid mcp for vscode

#### Install npm, uv on windows
  * https://wiki.visrc.com/doku.php?id=windows:winadmin:npm_user

#### Setup memvid mcp server

* https://github.com/khgs2411/memvid_mcp

Running local mpx server may not work due to not available PATH for non admin user.

Install memvid-mcp-server locally

If you have Node.js installed but npx isn't in PATH, install the package globally or locally:

npm install -g memvid-mcp-server

Then update the ~/.config/kilo.json (~/.config/kilo.jsonc for global settings) configuration to use the direct path:

```
{
  "command": ["memvid-mcp-server"],
  ...
}

```

Example:

```
"$schema": "https://app.kilo.ai/config.json",
  "mcp": {
    "memvid": {
      "type": "local",
      "command": [
        "npx",
        "-y",
        "memvid-mcp-server@latest"
      ],
      "environment": {
        "OPENAI_API_KEY": "sk-proj-qoGHEBwSHb41x8slxaziGjq1RSp5Sz3FT3BlbkFJYfp15vPobwEm0e-Ib0zdRhOTntpQsGf2qSY8kA3FR6T4np3tMlrc79yjyXLWtcpEd1BaE-IbkA",
        "MEMVID_LOCAL_STORAGE": "0"
      },
      
```

```
 "$schema": "https://app.kilo.ai/config.json",
  "mcp": {
    "memvid": {
      "type": "local",
      "command": ["memvid-mcp-server"],
      "environment": {
        "OPENAI_API_KEY": "sk-proj-NEZJMF3Yljd-lfp15vPobwEm0e-Ib0zdRhOTntpQsGf2qSY8kA3FR6T4np3tMlrc79yjyXLWtcpEd1BaE-IbkA",
        "MEMVID_LOCAL_STORAGE": "0"
      },
```

Note: If you do not provide OPENAI_API_KEY, the ask_memory tool will not be available, but you can still use create_or_open_memory, add_content, and search_memory (lexical mode).

MEMVID_LOCAL_STORAGE Set to "1" to store memory files in ./memvid_mcp (current directory) instead of ~/.memvid_mcp.

**This only work in external powershell not in VS Code sandbox. See troubleshooting**
If PATH only defined in non admin user $profile, use this command

["powershell", "-Command", ". $PROFILE; npx -y memvid-mcp-server@latest"]
     
#### Troubleshooting

VS Code use sandbox environment where non admin user $profile is not available.

##### Set PATH
     powershell -Command "[Environment]::SetEnvironmentVariable('Path', 'D:\devfs\python;D:\devfs\node-v24.15.0-win-x64;' + [Environment]::GetEnvironmentVariable('Path', 'User'), 'User')" 2>&1; echo "PATH_UPDATED:$?"
     
Edit PATH as non admin user:

rundll32 sysdm.cpl,EditEnvironmentVariables

Add full PATH:

D:\devfs\python;D:\devfs\node-v24.15.0-win-x64;D:\devfs\node-v24.15.0-win-x64;D:\devfs\node-v24.15.0-win-x64;C:\Users\13223\.local\bin;C:\Users\13223\AppData\Local\Microsoft\WindowsApps;C:\Users\13223\AppData\Local\Programs\Git\cmd;C:\Users\13223\AppData\Local\Programs\Microsoft VS Code Insiders\bin;C:\Users\13223\AppData\Local\Programs\DockerDesktop\resources\bin;D:\devfs\jdk-21.0.10+7\bin;D:\devfs\apache-maven-3.9.15\bin;D:\devfs\node-v24.15.0-win-x64;.local\bin;D:\devfs\vim92;
     
##### Set $env

[Environment]::SetEnvironmentVariable("OPENAI_API_KEY", "sk-your-key-here", "User")

Then restart your computer (not just VS Code) to ensure it propagates system-wide.

Check $env in terminal:
  
     $env:OPENAI_API_KEY

Alternative: Set it in VS Code's terminal settings If you need it available immediately in the integrated terminal regardless, add this to your VS Code settings.json:

"terminal.integrated.env.windows": {
    "OPENAI_API_KEY": "sk-your-key-here"
}

### pdf-reader mcp for vscode

* https://github.com/SylphxAI/pdf-reader-mcp

A PDF Reader MCP (Model Context Protocol) server is an integration that allows AI agents and IDEs (like Claude or VS Code) to securely read, search, and extract data from PDF files without token-heavy context dumping. It solves large PDF limitations by converting binary documents into structured, searchable evidence.Advanced implementations like SylphxAI/pdf-reader-mcp go beyond simple text extraction, acting as an "Agent Document Twin." They provide:Evidence-based querying: pinpoint exactly which paragraph, table, or bounding box proves an answer to avoid hallucinations.Vis-mapping: interpret layouts, images, charts, and hidden text through built-in OCR and visual candidate crops.For simpler, text-only extraction workflows, lighter MCP servers are available on the MCP Market, including the mcp-pdf-reader PyPI package and @sylphlab/pdf-reader-mcp. These typically support basic functionalities such as:Opening and closing PDFs using file paths or URLs.Listing metadata (e.g., page counts, title, author, creation dates).Extracting text by full document, page ranges, or specific search queries.

npm install -g @sylphx/pdf-reader-mcp

## PDF reader MCP
* https://lobehub.com/mcp/gstiebler-pdf-mcp-server

Edit summary

Note: By editing this page you agree to license your content under the following license:

Trace:

Page locked

You've loaded an old revision of the document! If you save it, you will create a new version with this data.

Navigation

Search

Toolbox

Page locked

**You've loaded an old revision of the document!** If you save it, you will create a new version with this data.

Navigation

Search

Toolbox

You've loaded an old revision of the document! If you save it, you will create a new version with this data.