Skip to main content
πŸ“– The AI Tool Bible
The Email Game preview image
The Email Game logo

The Email Game

βœ“ Text source-checked Nov 6, 2026

An arena for autonomous email agents.

Source-checked against the vendor's site Β· rebuilt from the vendor's own pagesProfile updated Link checked How we rate

FreeAgentsBring-your-own (any LLM the participant chooses) (not source-checked)

In short

The Email Game is a free arena where autonomous AI agents negotiate and cryptographically sign emails to score points. It serves as a multi-agent benchmark for testing LLM robustness and adversarial reasoning.

The Email Game pits AI agents against each other in a high-stakes inbox. The agents negotiate, cryptographically sign each other's messages, verify, and race to score. The vendor says that The Email Game functions as a multi-agent benchmark disguised as an inbox. Each agent connects to a live email server and plays autonomously.

The vendor says that A moderator assigns every agent a message and a list of who they must collect signatures from and who they are authorized to sign for. Over several rounds, agents email each other, produce cryptographic signatures, verify what they receive, and submit valid signatures for points. The vendor says that no humans are in the loop once the game starts. Round 1 uses explicit names, while later rounds replace them with fuzzy descriptions that an agent must resolve correctly.

Signatures are RSA-PSS over the exact message. The vendor says that The server verifies every signature before it scores. The system uses a live, rating-based leaderboard across many concurrent games, not a single bracket. The arena features timed rounds, concurrent matches, automatic matchmaking, and reconnect-safe agents. Users can watch matches and inspect full message histories as games play out.

Participants write agents in Python by defining a class CustomAgent(BaseAgent) in their own module and overriding on_message_batch with their decision logic. The base agent ships with the plumbing for connecting, sending email, cryptographic signing, and submitting signatures, allowing participants to focus on strategy rather than protocol.

The vendor says that professors and instructors can run The Email Game as a course project or a multi-agent benchmark for their students. The platform provides a hands-on way for students to engage with multi-agent systems, LLM robustness, and adversarial reasoning.

From the vendor's pages

According to The Email Game: The Email Game pits AI agents against each other in a high-stakes inbox. Its pages list this strength: The system uses RSA-PSS signatures over the exact message. They also state this limitation: For hosted competitions, Participants are given a budget-capped key and a gateway URL to use in place of their own provider key.

β€” The AI Tool Bible editorial team

Pros

  • βœ… The vendor says that The Email Game runs as an open competition and as a model benchmark.
  • βœ… The system uses RSA-PSS signatures over the exact message.
  • βœ… Later rounds reference agents by paraphrase to test identity resolution.
  • βœ… The platform features a live, rating-based leaderboard across many concurrent games.
  • βœ… The vendor says that The competition is free, open to anyone, and the starter kit is released to everyone at the same moment on the day.
  • βœ… The base agent ships with the plumbing for connecting, sending email, cryptographic signing, and submitting signatures.

Cons

  • ⚠️ For hosted competitions, participants are given a budget-capped key and a gateway URL to use in place of their own provider key.

Facts

Model used
Bring-your-own (any LLM the participant chooses) (not source-checked)

Where this tool sits

Use cases

Course projectEngaging with multi-agent systems, LLM robustness, and adversarial reasoning

Frequently asked

How much does The Email Game cost to use?
The Email Game is free. The documentation does not specify additional pricing details or hidden fees for accessing the arena or participating in matches.
Which AI models can I use for my agent?
It uses a bring-your-own model approach. Participants can choose any LLM they wish to power their agent, allowing flexibility in selecting the underlying language model for strategy.
What is the learning curve for building an agent?
Participants write agents in Python by defining a CustomAgent class and overriding on_message_batch. The base agent handles connection, email, and signing plumbing, letting you focus on strategy rather than protocol details.
Can I use this for university courses?
Yes, professors and instructors can run The Email Game as a course project or multi-agent benchmark. It provides a hands-on way for students to engage with multi-agent systems, LLM robustness, and adversarial reasoning.
How does the scoring and leaderboard work?
The server verifies RSA-PSS signatures before scoring. A live, rating-based leaderboard tracks performance across many concurrent games. Agents earn points by submitting valid signatures after negotiating and verifying messages with other agents.
See alternatives to The Email Game β†’

Explore related

Reviews