Lanes Link
One open-source MCP endpoint you run yourself. Make a connection once, then reach it from every agent, behind one set of permissions.
Getting Started
Built-ins
Built-ins
Memory, tasks, assets, skills, identity, entities, the vault, and the audit log. Your own material, with no account to connect and no credential to hold.
Memory
What is true, kept as plain Markdown and served back to every later session. How to write it, how to search it, and why you may want it read-only.
Tasks
What is to be done, with a status so it can be closed. Six statuses, why list shows three, and how tasks differ from memory.
Assets
A file kept under its own name, with no id, prefix, sidecar, or index. What the naming rules are, and why an asset carries no description.
Skills
Your own procedures, surfaced as MCP prompts rather than tools, so a person selects one and a model cannot choose it for you.
Vault
Credentials your agents can spend without reading. One capability per item, no listing, and why a value written now is not readable until the endpoint restarts.
Identity
Who you are, so an agent writes as you rather than as itself. Read-only by construction, and the one built-in with no write path at all.
Entities
Who everyone else is: the people, companies and projects you deal with, and how to reach each of them. Looked up rather than remembered, and never guessed.
Audit log
Every capability call, allowed or refused, in a hash-chained append-only record. What an entry keeps, what it withholds, and how to verify nothing was altered.
Connections
Connections
What a connection is, the methods behind one, and what happens the moment you make it. The entry point to the provider catalogue and to narrowing what an agent may do.
Scopes and permissions
The two different things called scope, how to narrow one instantly and widen the other, and what can and cannot be set per connection.
Add your own
Any MCP server, REST API, IMAP mailbox, CalDAV or CardDAV server, or folder on your machine
Providers
107 more in the catalogue.
Reference
Under the Hood
Architecture
How it fits together, and the dispatch path a call takes
Security model
The threat model and the guarantee table
Capabilities
Why each capability is a tool, a resource, or a prompt, where its list comes from, and what that decides about what can be checked.
Connectivity coverage
Which connectivity and credential types compose, which are closed, and what none of them covers
Deploying to Cloud Run
Cloud Run in full: cold starts, scaling, IAM, the image
Registering your own Google client
The console walkthrough, service account keys, and the Workspace MCP servers
Google verification: what goes in the console
The scope justifications Google's review asks for, and why each scope is the narrowest that works