Connect coding agents to your Althea and, through her, a consent-first network of ML researchers.
claude mcp add althea-mcp -- uvx althea-mcp{
"mcpServers": {
"althea-mcp": {
"command": "uvx",
"args": ["althea-mcp"],
"env": {
"ALTHEA_THREAD_KEY": "<althea_thread_key>"
}
}
}
}ALTHEA_THREAD_KEYMCP Servers overview
<!-- mcp-name: io.github.tiptreesystems/althea-mcp -->
<p align="center">
<img src="https://raw.githubusercontent.com/tiptreesystems/althea-mcp/main/assets/althea-logo.svg" alt="Althea" width="156" />
</p>
<h1 align="center">Althea MCP</h1>
<p align="center">
<strong>A direct line from your coding agent to your Althea.</strong><br />
Shared research memory and a consent-first network of verified ML researchers.
</p>
<p align="center">
<a href="https://github.com/tiptreesystems/althea-mcp/actions/workflows/ci.yml"><img src="https://github.com/tiptreesystems/althea-mcp/actions/workflows/ci.yml/badge.svg" alt="CI status" /></a>
<a href="https://github.com/tiptreesystems/althea-mcp/releases/latest"><img src="https://img.shields.io/github/v/release/tiptreesystems/althea-mcp?label=Release" alt="Latest GitHub release" /></a>
<a href="https://pypi.org/project/althea-mcp/"><img src="https://img.shields.io/pypi/v/althea-mcp?label=PyPI" alt="PyPI version" /></a>
<a href="https://www.python.org/"><img src="https://img.shields.io/badge/Python-3.11%2B-3776AB?logo=python&logoColor=white" alt="Python 3.11+" /></a>
<a href="https://registry.modelcontextprotocol.io/?q=io.github.tiptreesystems%2Falthea-mcp"><img src="https://img.shields.io/badge/MCP%20Registry-listed-638B8D" alt="MCP Registry listing" /></a>
<a href="https://github.com/tiptreesystems/althea-mcp/blob/main/LICENSE"><img src="https://img.shields.io/badge/license-MIT-2F6D70.svg" alt="MIT license" /></a>
</p>
Give Codex, Claude Code, and other coding agents a direct line to your personal
[Althea](https://althea.tiptreesystems.com). They can hand her questions, code,
and context from the repository in front of them. Althea brings your shared
research memory and, when useful, can ask a consent-first network of verified
ML researchers for help.
- **Agents that can talk to each other.** Pass questions, code, experiment
results, and repository context between your coding agent and Althea.
- **Your Althea, with your context.** MCP uses your existing account, profile,
and long-term research memory.
- **A route into the researcher network.** Althea can ask verified ML
researchers for help, with consent before private context is shared.
- **Client-specific conversations.** Codex and Claude stay separate when you
give them distinct thread keys, while Althea keeps your shared context.
- **Work asynchronously.** Wait for a reply, leave a message for longer work,
or read the thread later.
[Setup](#setup) · [Connect a client](#connect-an-mcp-client) ·
[First conversation](#start-a-conversation) · [Tools](#tools) ·
[Security and privacy](#security-and-privacy) ·
[Development](#local-development)
## Setup
You need access to Althea and [uv](https://docs.astral.sh/uv/getting-started/installation/).
Install Althea MCP from PyPI, then sign in with an emailed verification code:
```bash
uv tool install althea-mcp
althea-mcp setup
```
If the install succeeds but the command is not found, run `uv tool update-shell`
and open a new terminal.
Run setup yourself in an interactive terminal. Enter your email and verification
code there, not in an AI chat.
```text
Connect your personal Althea
-----------------------------
Email: you@example.com
A verification code was sent to you@example.com.
Verification code:
Althea MCP is ready.
Credentials saved to ~/.config/althea-mcp/credentials.json
```
Setup follows the same account journey as the Althea web app:
- Existing accounts receive a verification code by email.
- Eligible new accounts are asked for a name, then receive a verification code.
- Accounts that still need access are sent to the
[Althea application](https://tiptreesystems.com/apply). Complete it, then run
setup again.
The installed server starts without fetching code on every launch. Its tools
still need a network connection to Althea. Upgrade the package later with:
```bash
uv tool upgrade althea-mcp
```
## Connect an MCP client
Complete setup once, then add the local stdio server. Give each client a
different thread key.
> [!CAUTION]
> This server can read private Althea history and send real messages. Register
> it only in clients and projects you trust.
### Codex
```bash
codex mcp add althea \
--env ALTHEA_THREAD_KEY=codex \
-- althea-mcp
codex mcp list
```
`codex mcp add` writes to your user configuration. Equivalent
`~/.codex/config.toml`:
```toml
[mcp_servers.althea]
command = "althea-mcp"
[mcp_servers.althea.env]
ALTHEA_THREAD_KEY = "codex"
```
### Claude Code
```bash
claude mcp add althea \
-e ALTHEA_THREAD_KEY=claude-code \
-- althea-mcp
claude mcp list
```
That command uses Claude Code's project-local scope. Add `--scope user`
immediately before `althea` only if you want the server available in every
project you open.
A user-wide manual entry under the top-level `mcpServers` object in
`~/.claude.json` looks like this:
```json
{
"mcpServers": {
"althea": {
"type": "stdio",
"command": "althea-mcp",
"env": {
"ALTHEA_THREAD_KEY": "claude-code"
}
}
}
}
```
### Claude Desktop and other local clients
Find the installed command:
```bash
command -v althea-mcp
```
On Windows PowerShell:
```powershell
(Get-Command althea-mcp).Source
```
GUI applications sometimes receive a smaller `PATH` than your terminal. If
`althea-mcp` is not found, use the absolute path returned above:
```json
{
"mcpServers": {
"althea": {
"command": "/absolute/path/to/althea-mcp",
"env": {
"ALTHEA_THREAD_KEY": "claude-desktop"
}
}
}
}
```
Restart the client after changing its MCP configuration.
## Start a conversation
Try one of these prompts in your MCP client:
```text
Ask my Althea whether anyone in the researcher network has worked on this
failure mode. Share only the sanitized summary below:
[summary]
```
```text
Send this context to my Althea without waiting for a reply:
The latest experiment rules out the data-loader hypothesis. The remaining
failure starts after gradient accumulation.
```
```text
Check whether my Althea has replied in the MCP thread.
```
Messages continue the configured thread. They are real messages to your
personal Althea and may cause her to begin work.
## Tools
| Tool | What it does |
| --- | --- |
| `ask_althea(message)` | Sends a message and uses a 120-second polling window by default for the first reply. If the window expires, the work may still continue. |
| `send_message_to_althea(message)` | Sends a message immediately and returns a receipt without waiting. Use it for context, notes, and longer requests. |
| `get_althea_messages(sender=None, limit=10)` | Returns 1 to 100 recent messages in chronological order. Optionally filter by `user`, `assistant`, or `system`. |
The two send tools are marked as state-changing. Message retrieval is read-only.
The current tool surface is text-only. It does not expose attachments or other
Althea artifacts.
## How it works
```text
Codex / Claude Code / another coding agent
│ local stdio
▼
Althea MCP
│ HTTPS
▼
your Althea
│
├── your profile and long-term research memory
├── a persisted conversation for this thread key
└── a consent-first network of verified ML researchers
```
Althea MCP is a small public adapter. The frontend authenticates the user,
resolves their canonical Althea, sends messages, and stores the resulting
conversation. The package polls those persisted messages when a caller waits
for a reply. The coding agent talks to Althea, and Althea coordinates with the
researcher network when a request would benefit from outside expertise.
## Sign-in and sessions
Setup creates a normal Althea usage session:
- The access session lasts 7 days.
- Its rotating refresh-token family lasts 14 days from sign-in.
- Access tokens refresh automatically before expiry.
- A cross-process lock lets several local MCP clients share the credential file
without racing refresh-token rotation.
- When the session expires or is revoked, run `althea-mcp setup` again in a
separate terminal. Running MCP processes reload the replacement credentials.
Credentials are stored as plaintext JSON at
`~/.config/althea-mcp/credentials.json`. On operating systems that support Unix
file modes, the credential file is set to `0600`. A parent directory created by
Althea MCP is requested as `0700`; check the permissions yourself when using a
pre-existing custom directory.
## Security and privacy
- Messages sent through these tools are persisted in your Althea account, like
messages from Althea's other channels.
- The local stdio process sends messages and credentials to Althea over HTTPS.
Even with local transport, conversation data leaves your machine.
- The connected MCP host and model can see tool arguments and replies. They may
retain that data under their own privacy and retention policies.
- When Althea asks the researcher network for help, she requests consent before
sharing information drawn from a private conversation.
- The package sends access credentials only to the Althea origin they were
issued for. Changing `ALTHEA_APP_URL` requires a new setup.
- The refresh token is sent only to the session rotation endpoint.
- Email verification codes and the credential file are secrets. Never paste
them into an AI chat, terminal transcript, issue, or pull request.
- To remove the local session, close MCP clients and delete the credential file.
This forgets the session on that machine. Contact Tiptree if a copied session
must also be revoked.
Please report vulnerabilities using
[the security policy](https://github.com/tiptreesystems/althea-mcp/security/policy).
## Configuration
Most users only need `althea-mcp setup` and a client-specific thread key.
| Environment variable | Default | Purpose |What people ask about althea-mcp
What is tiptreesystems/althea-mcp?
+
tiptreesystems/althea-mcp is mcp servers for the Claude AI ecosystem. Connect coding agents to your Althea and, through her, a consent-first network of ML researchers. It has 0 GitHub stars and was last updated today.
How do I install althea-mcp?
+
You can install althea-mcp by cloning the repository (https://github.com/tiptreesystems/althea-mcp) or following the README instructions on GitHub. ClaudeWave also provides quick install blocks on this page.
Is tiptreesystems/althea-mcp safe to use?
+
tiptreesystems/althea-mcp has not been audited yet by our security agent. Review the original repository on GitHub before using it in production.
Who maintains tiptreesystems/althea-mcp?
+
tiptreesystems/althea-mcp is maintained by tiptreesystems. The last recorded GitHub activity is from today, with 1 open issues.
Are there alternatives to althea-mcp?
+
Yes. On ClaudeWave you can browse similar mcp servers at /categories/mcp, sorted by popularity or recent activity.
Deploy althea-mcp to your cloud
Ship this repo to production in minutes. Each platform spins up its own environment with editable env vars.
Maintain this repo? Add a badge to your README
Drop the badge into your GitHub README to show it's tracked on ClaudeWave. Each badge links back to this page and reflects the live Trust Score.
[](https://claudewave.com/repo/tiptreesystems-althea-mcp)<a href="https://claudewave.com/repo/tiptreesystems-althea-mcp"><img src="https://claudewave.com/api/badge/tiptreesystems-althea-mcp" alt="Featured on ClaudeWave: tiptreesystems/althea-mcp" width="320" height="64" /></a>More MCP Servers
Fair-code workflow automation platform with native AI capabilities. Combine visual building with custom code, self-host or cloud, 400+ integrations.
User-friendly AI Interface (Supports Ollama, OpenAI API, ...)
An open-source AI agent that brings the power of Gemini directly into your terminal.
The fastest path to AI-powered full stack observability, even for lean teams.
Real-time global intelligence dashboard. AI-powered news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational awareness interface
🕷️ An adaptive Web Scraping framework that handles everything from a single request to a full-scale crawl!