Engram by Thomas Sprock
Clip the active tab to Markdown and save it to your Engram server.
Extension Metadata
Screenshots
About this extension
Engram — Browser Extension for Your Self-Hosted Knowledge Vault
Your knowledge. Your server. Your rules.
Most tools that give an AI assistant a "memory" do so by routing your notes through someone else's infrastructure. Engram takes the opposite position: you run the server, you hold the data. This browser extension connects Chrome to your self-hosted Engram instance — letting you clip any web page into your personal knowledge vault with a single click, from wherever you are on the web.
Nothing leaves your infrastructure except what you explicitly save. No account required. No telemetry. No third-party cloud in the middle.
WHAT THE EXTENSION DOES
When you find something worth keeping — an article, a documentation page, a research result, a product spec — the Engram extension sends it directly to your Engram server. The server fetches the full page content, stores it as a plain Markdown file with YAML frontmatter, and makes it immediately searchable across all three surfaces the server exposes: the web UI, the REST API, and the MCP endpoint your LLM client connects to.
That last part is the point. Once a page is in your vault, your AI assistant can read it, reference it, and reason over it — because Engram runs a spec-compliant MCP server. You clip something in Chrome; your assistant knows about it in the next conversation. No copy-paste, no manual uploads, no context window workarounds.
HOW IT FITS INTO THE ENGRAM ARCHITECTURE
Engram is a single backend process that serves three surfaces simultaneously:
The extension authenticates against your server using a bearer token you set yourself. It talks only to the server you configure — there is no relay, no intermediary, no Engram-operated cloud service. If your server is unreachable, the extension cannot send data anywhere. That constraint is intentional.
Your vault is a directory of plain Markdown files on disk. You can read it, back it up, version it with Git, or move it to a different machine with standard tools — whether or not Engram is running. The server maintains a SQLite FTS5 index for fast full-text search, but the index is derived from the files, not the other way around. The files are always the source of truth.
BEFORE YOU INSTALL
This extension requires a running Engram server. Engram is open-source (MIT licensed) and self-hosted — you run it on your own hardware or a VPS. The quick-start uses Docker Compose and takes a few minutes. A static demo of the web UI is available at https://t11z.github.io/engram/demo/ if you want to see the interface before committing to the setup.
Full installation instructions, configuration reference, MCP tool reference, and API documentation live at: https://t11z.github.io/engram/
Source code: https://github.com/t11z/engram
SETUP IN THREE STEPS
That's the entire flow. No OAuth dance, no account creation, no email address.
PRIVACY
The extension sends data to exactly one destination: the server URL you configure in the options page. It does not communicate with any Engram-operated server, analytics service, or third-party endpoint. It requests only the permissions necessary to read the current tab's URL and title, and to make authenticated requests to your configured server.
Because you host the server yourself, you control where the data is stored, how it is backed up, and who can access it. The single-user, self-hosted design is not a limitation — it is the feature.
COMPATIBILITY
OPEN SOURCE
Engram is MIT licensed. Contributions, bug reports, and feature requests are welcome at https://github.com/t11z/engram.
Your knowledge. Your server. Your rules.
Most tools that give an AI assistant a "memory" do so by routing your notes through someone else's infrastructure. Engram takes the opposite position: you run the server, you hold the data. This browser extension connects Chrome to your self-hosted Engram instance — letting you clip any web page into your personal knowledge vault with a single click, from wherever you are on the web.
Nothing leaves your infrastructure except what you explicitly save. No account required. No telemetry. No third-party cloud in the middle.
WHAT THE EXTENSION DOES
When you find something worth keeping — an article, a documentation page, a research result, a product spec — the Engram extension sends it directly to your Engram server. The server fetches the full page content, stores it as a plain Markdown file with YAML frontmatter, and makes it immediately searchable across all three surfaces the server exposes: the web UI, the REST API, and the MCP endpoint your LLM client connects to.
That last part is the point. Once a page is in your vault, your AI assistant can read it, reference it, and reason over it — because Engram runs a spec-compliant MCP server. You clip something in Chrome; your assistant knows about it in the next conversation. No copy-paste, no manual uploads, no context window workarounds.
HOW IT FITS INTO THE ENGRAM ARCHITECTURE
Engram is a single backend process that serves three surfaces simultaneously:
- A spec-compliant MCP endpoint for LLM clients (tested primarily with Claude)
- A versioned REST API — which this extension uses
- A lightweight web UI for browsing, reading, and managing your vault
The extension authenticates against your server using a bearer token you set yourself. It talks only to the server you configure — there is no relay, no intermediary, no Engram-operated cloud service. If your server is unreachable, the extension cannot send data anywhere. That constraint is intentional.
Your vault is a directory of plain Markdown files on disk. You can read it, back it up, version it with Git, or move it to a different machine with standard tools — whether or not Engram is running. The server maintains a SQLite FTS5 index for fast full-text search, but the index is derived from the files, not the other way around. The files are always the source of truth.
BEFORE YOU INSTALL
This extension requires a running Engram server. Engram is open-source (MIT licensed) and self-hosted — you run it on your own hardware or a VPS. The quick-start uses Docker Compose and takes a few minutes. A static demo of the web UI is available at https://t11z.github.io/engram/demo/ if you want to see the interface before committing to the setup.
Full installation instructions, configuration reference, MCP tool reference, and API documentation live at: https://t11z.github.io/engram/
Source code: https://github.com/t11z/engram
SETUP IN THREE STEPS
- Deploy your Engram server (Docker Compose, ~5 minutes — see the installation guide)
- Install this extension and enter your server URL and bearer token in the options page
- Click the Engram icon on any page to save it to your vault
That's the entire flow. No OAuth dance, no account creation, no email address.
PRIVACY
The extension sends data to exactly one destination: the server URL you configure in the options page. It does not communicate with any Engram-operated server, analytics service, or third-party endpoint. It requests only the permissions necessary to read the current tab's URL and title, and to make authenticated requests to your configured server.
Because you host the server yourself, you control where the data is stored, how it is backed up, and who can access it. The single-user, self-hosted design is not a limitation — it is the feature.
COMPATIBILITY
- Requires a running Engram server (open-source, self-hosted)
- MCP client compatibility: any spec-compliant MCP client; tested primarily with Claude
- iOS Shortcut available for clipping from Safari on iPhone (see the documentation)
OPEN SOURCE
Engram is MIT licensed. Contributions, bug reports, and feature requests are welcome at https://github.com/t11z/engram.
Rated 0 by 0 reviewers
Permissions and data
More information
- Add-on Links
- Version
- 2.0.2
- Size
- 122.39 KB
- Last updated
- 20 days ago (Jun 3, 2026)
- Related Categories
- License
- MIT License
- Privacy Policy
- Read the privacy policy for this add-on
- Version History
- Add to collection