Enhanced GeoGebra MCP
A local stdio MCP server for a self-hosted GeoGebra 5.0 Web bundle. It enables executing GeoGebra commands with structured feedback, querying commands, evaluating CAS, managing state, and exporting images using headless Chromium.
README
Enhanced GeoGebra MCP
Local stdio MCP server for a self-hosted GeoGebra Web bundle.
enhanced-geogebra-mcp wraps a modified GeoGebra 5.0 Web bundle under geogebra/. At runtime it starts a localhost-only static asset server, launches headless Chromium with Playwright, loads public/harness.html, obtains the GeoGebra JavaScript API through GGBApplet.appletOnLoad, and exposes a stdio MCP server.
What Is Enhanced
The bundled applet is expected to expose evalCommandResult(command), a modified API added to the GeoGebra runtime. ggb_eval_command uses that method directly.
Plain evalCommand only reports whether a command was accepted. evalCommandResult returns structured command feedback:
ok: whether the command succeeded.result: command result text returned by the applet.labels: labels created by the command.error: command error text when execution fails.objectCount: construction object count after execution.apiMethod: the runtime method used by the server.
There is no fallback to evalCommand. If evalCommandResult is not present in the loaded bundle, command evaluation fails.
Command Feedback Flow
flowchart LR
A["Choose construction command"] --> B["ggb_query_commands<br/>GeoGebra 5.0 reference"]
B --> C["ggb_eval_command<br/>call evalCommandResult"]
C --> D{"Command feedback"}
D -->|"ok"| E["Use labels/result"]
D -->|"error"| F["Revise command"]
F --> B
E --> G["ggb_get_objects / ggb_export_image<br/>verify construction"]
ReAct Comparison
The main difference is where command observation is produced. A conventional GeoGebra MCP wrapper built around evalCommand usually needs a follow-up inspection step to infer what changed. This package returns command-level observation from ggb_eval_command itself.
For LLM-driven construction loops, the practical advantages are:
- Command execution results add useful context immediately after each action.
- Failed commands return error text, giving the model a concrete correction signal and reducing unsupported guesses.
flowchart TD
subgraph T["Conventional GeoGebra MCP"]
direction LR
T1["Thought<br/>choose command"] --> T2["Action<br/>evalCommand"]
T2 --> T3["Limited feedback<br/>accepted / rejected"]
T3 --> T4["Extra inspection<br/>objects / XML / image"]
T4 --> T5["Infer result, labels, or error"]
end
T5 --> V["VS"]
subgraph E["enhanced-geogebra-mcp"]
direction LR
E1["Thought<br/>choose command"] --> E2["Action<br/>ggb_eval_command"]
E2 --> E3["Execution feedback<br/>ok + result + labels"]
E3 --> E4["Error feedback<br/>concrete message on failure"]
E4 --> E5["Revise command with more context"]
end
V --> E1
GeoGebra 5.0 Scope
The bundled runtime is GeoGebra 5.0. Commands passed to ggb_eval_command should use GeoGebra 5.0 command names, syntax, and signatures. Use ggb_query_commands when command syntax is uncertain.
Requirements
- Node.js 20 or newer.
- npm or npx.
- An MCP client that supports stdio servers.
- Chromium, either installed by Playwright or provided through an existing Chrome/Chromium executable.
- The modified GeoGebra 5.0 runtime that exposes
evalCommandResult. The npm package includes this runtime undergeogebra/.
Distribution
This package is released under the MIT license and is intended for npm distribution as enhanced-geogebra-mcp.
This project is not affiliated with GeoGebra. The bundled runtime targets GeoGebra 5.0; command syntax, command reference data, and runtime behavior should stay aligned with that version.
Security And Network
- The MCP server communicates over stdio.
- GeoGebra assets are served from a localhost-only static asset server.
- Browser execution is handled by Playwright and Chromium.
- Browser requests to non-local URLs are blocked by default.
--allow-external-networkexplicitly allows non-local browser requests.- Large state payloads can be returned as MCP resources when they exceed
--max-inline-chars.
Known Limitations
- Command syntax is scoped to GeoGebra 5.0.
- Commands should use English GeoGebra input-bar names.
- The server needs a working Chromium runtime.
Versioning And Compatibility
- The npm package version tracks the MCP wrapper.
- The bundled GeoGebra runtime is currently GeoGebra 5.0.
- The bundled command reference is expected to match GeoGebra 5.0 syntax.
- Runtime or API changes that affect
evalCommandResultshould be documented inCHANGELOG.md.
Install From npm
After the package is published to npm, it can be used directly with npx:
npx -y enhanced-geogebra-mcp --help
MCP client config example:
{
"mcpServers": {
"enhanced-geogebra": {
"command": "npx",
"args": ["-y", "enhanced-geogebra-mcp"]
}
}
}
It can also be installed globally:
npm install -g enhanced-geogebra-mcp
enhanced-geogebra-mcp --help
Global install config example:
{
"mcpServers": {
"enhanced-geogebra": {
"command": "enhanced-geogebra-mcp",
"args": []
}
}
}
The server uses Playwright to launch Chromium. If no compatible Chromium is available, install one:
npx playwright install chromium
Or point the server at an existing Chrome/Chromium executable with GEOGEBRA_MCP_BROWSER_EXECUTABLE or --browser-executable-path.
Install From Source
npm install
npx playwright install chromium
npm run build
Run As MCP Stdio Server
npm run build
node dist/index.js
Client config example:
{
"mcpServers": {
"enhanced-geogebra": {
"command": "node",
"args": ["/absolute/path/to/enhanced-geogebra-mcp/dist/index.js"]
}
}
}
After packaging:
npm run pack:local
The package exposes an enhanced-geogebra-mcp binary.
Tools
ggb_create_session: create a headless GeoGebra applet session.ggb_eval_command: evaluate English GeoGebra input-bar commands through the bundled applet'sevalCommandResultAPI. The payload includesok,result,labels,error,objectCount, andapiMethod. This package expects that modified API to exist.ggb_query_commands: search the bundled GeoGebra command reference by name, syntax, examples, tags, or scope.ggb_eval_cas: evaluate a CAS expression.ggb_get_objects: list construction objects and compact metadata.ggb_get_state: return XML, base64.ggb, or fileJSON state.ggb_set_state: load XML, base64.ggb, or fileJSON state.ggb_export_image: export PNG, SVG, or screenshot.ggb_set_view: set common view options.ggb_reset: clear or reset the construction.ggb_close_session: close a browser-backed session.ggb_list_sessions: list active sessions.
Options
enhanced-geogebra-mcp --help
Notable options:
--geogebra-root <path>: use a different GeoGebra bundle root.--asset-port <port>: pin the internal localhost asset server port.--browser-executable-path <path>: use an existing Chrome/Chromium executable.--allow-external-network: allow the browser to load non-local URLs. Default is blocked.--max-inline-chars <count>: return larger payloads as MCP resources.
You can also set GEOGEBRA_MCP_BROWSER_EXECUTABLE=/path/to/chrome-or-chromium when Playwright's bundled Chromium is not installed.
Smoke Checks
npm run smoke:harness
npm run smoke:mcp
smoke:harness tests the local asset server plus browser runtime directly. smoke:mcp builds the package, starts the stdio MCP server as a child process, and calls the core tools through an MCP client.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.