About Glasser
Glasser provides a hosted MCP server and agent plugins for accessing paid data APIs through a single account. It is for agents and their users who need to find endpoints, review pricing, and run data tasks with charges drawn from a selected Workspace balance.
Key features
- 01Search and inspect API catalog and prices
- 02Run data tasks and view past runs
- 03Check balance and stop runs
- 04OAuth sign-in or key-based access
- 05Optional CLI for scripting and CI
Use cases
- 01Find an API endpoint for a data task
- 02Review endpoint pricing before execution
- 03Run enrichment, email verification, or SEO tasks
- 04Use the CLI for scripting or CI results
MCP support
The hosted MCP server is available at https://api.glasser.ai/mcp. Its tools search and inspect the catalog, run tasks, retrieve or stop runs, and check the Workspace balance.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://api.glasser.ai/mcp?utm_source=mcp-registry&utm_medium=mcp_directoryclaude mcp add --transport http glasser https://api.glasser.ai/mcp?utm_source=mcp-registry&utm_medium=mcp_directory{"mcpServers":{"glasser":{"type":"http","url":"https://api.glasser.ai/mcp?utm_source=mcp-registry&utm_medium=mcp_directory"}}}Open in Cursor{"servers":{"glasser":{"type":"http","url":"https://api.glasser.ai/mcp?utm_source=mcp-registry&utm_medium=mcp_directory"}}}Open in VS Code{"mcpServers":{"glasser":{"type":"http","url":"https://api.glasser.ai/mcp?utm_source=mcp-registry&utm_medium=mcp_directory"}}}{"mcpServers":{"glasser":{"serverUrl":"https://api.glasser.ai/mcp?utm_source=mcp-registry&utm_medium=mcp_directory"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project READMEGlasser plugins
One key, no contract, pay per call, across paid APIs bought wholesale.
Official agent plugins for Glasser. This repository is a distribution package: configuration plus markdown. It ships no executable code and no install script.
Works with Cursor, Claude Code, Grok Build, and any client that reads the Agent Plugins format.
Install
Cursor
Search glasser under Customize → Plugins, or open the listing on the
Cursor Marketplace. The first call opens
your browser to sign in.
Claude Code
/plugin marketplace add glasser-ai/plugins
/plugin install glasser@glasser
Any supported agent
npx plugins add glasser-ai/plugins
Detects every agent installed on the machine and installs to all of them.
Grok Build
/marketplace inside Grok Build once the plugin is listed. Until then, add
this repository under [[marketplace.sources]] in ~/.grok/config.toml; the
Grok Build docs have the syntax.
MCP only
Any MCP client that signs in with OAuth needs just the URL:
https://api.glasser.ai/mcp
Clients that cannot open a browser can send a Key instead — client-by-client snippets: https://glasser.ai/docs/mcp-server
Authentication
Nothing to create or paste. The first tool call opens your browser: sign in
to Glasser, choose the Workspace whose balance the agent may spend, and click
Allow. That mints a Key named MCP · <client> for that Workspace — it is
listed at https://app.glasser.ai/keys like any other Key, and revoking it
there disconnects the client. Disconnecting inside the client does not
revoke the Key; the next sign-in reuses it.
The optional CLI authenticates on its own with glasser login (browser
device authorization). That stores a Key for the CLI only.
What you get
MCP tools — search, inspect, run, runs_get, runs_list,
runs_stop, balance. Only run creates a Run and charges the Workspace
balance; the other tools read the catalog, past Runs and the Balance. Every
Run reports its own charge.
Skill — the workflow the agent follows: search the catalog, inspect an Endpoint's Price and charge clauses before running, run, then report the result and the charge. It uses the MCP tools by default and the CLI for results too large for the context window, for scripting and for CI.
Rule (Cursor only) — always-on spending guardrails.
Optional CLI — npm install -g @glasser-ai/cli. Same Key, same
Workspace, same Balance. Docs: https://glasser.ai/docs/cli
Network endpoints and credentials
For reviewers and anyone auditing what this plugin reaches.
| Endpoint | Used by | Credential |
|---|---|---|
https://api.glasser.ai/mcp | the MCP server config | the OAuth access token the client obtains at sign-in; the server resolves it to the MCP · <client> Key |
https://app.glasser.ai/api/auth | the client, at sign-in | OAuth authorization server: discovery, client registration, tokens |
https://app.glasser.ai | sign-in and the Workspace consent page; Key management; glasser login approval | your browser session |
https://api.glasser.ai/v1 | the optional CLI | Key stored by glasser login, or GLASSER_API_KEY |
https://registry.npmjs.org | the CLI's at-most-daily update check | none |
Nothing in this repository executes. The plugin sends no telemetry of its own. The service behind these endpoints is governed by Glasser's terms of service and privacy policy.
Repository layout
.cursor-plugin/marketplace.json Cursor catalog (one entry)
.claude-plugin/marketplace.json Claude Code catalog (one entry)
plugins/glasser/
├── plugin.json Agent Plugins manifest
├── .cursor-plugin/plugin.json Cursor manifest (carries rules/)
├── .claude-plugin/plugin.json Claude Code / Grok Build manifest
├── openclaw.plugin.json OpenClaw manifest (ClawHub requires it to publish)
├── mcp.json MCP server (read by Cursor, Agent Plugins)
├── .mcp.json identical copy (read by Claude Code, Grok Build)
├── skills/glasser/SKILL.md
├── rules/glasser-spending.mdc
├── assets/logo.svg 1:1, transparent (see DISTRIBUTION.md)
└── assets/icon.png the same mark, rasterised for ClawHub
scripts/ validators and the version bump
Releasing: see DISTRIBUTION.md.
License
MIT — see LICENSE. The Glasser service has its own terms.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.