The World Runs on Promises.
We Made Them Verifiable.
PROMISE, powered by the Agency Protocol, introduces a new architecture for accountability - a way for credible actors to demonstrate their promise-keeping ability in any domain.
Agency Protocol is free and open source software
Released under the MIT LicenseThe Current Web
Gameable Credibility
We rely on vague, easily gamed signals like 5-star ratings and anonymous reviews. This creates a world of information asymmetry where it's impossible to distinguish true capability from clever marketing.
The New Standard
Trust as an Asset
We use a formal grammar for commitment. Verifiable promises, economic stakes, and domain-specific merit create a high-fidelity trust system where accountability is the default and excellence is provable.
The Lifecycle of Trust
Every commitment on PROMISE is an observable process, transforming an intention into earned, verifiable merit.
1. Promise
An agent makes a specific, cryptographically signed commitment in a defined domain.
2. Stake
To prove seriousness, the agent locks a security deposit. Talk is no longer cheap.
3. Evidence
The agent provides tamper-proof evidence of fulfillment, from data logs to documents.
4. Merit
Based on assessment, the agent's stake is returned and their Merit score in that domain increases.
From Search to Certainty
Promise Fulfillment: An Evolving Story
The Search Engine
Search engines incentivize endless searching and clicking - not finding
The Promise Engine
Built for commitment. Its goal is to find a single, verifiable promise that matches your intention precisely. Behind every search is a promise that someone wants kept.
Matching Promises for "Quit smoking"
Assistants for the selected promise
Smoking Cessation Specialist · Merit: 99.2% · 142 promises kept · Boston, MA
Behavioral Therapist · Merit: 97.8% · 89 promises kept · Austin, TX
A System that makes good faith the winning strategy.
Rigorously validated by
mathematical proofs, agent-based models, and domain experts.
Honest Behavior is Optimal
GIVE TRUTH TELLERS THE ADVANTAGE.
The protocol is engineered so that keeping your promise is the most profitable long-term strategy. The economic incentives align self-interest with integrity.
ASSESSMENTS:
From `Simulation & Analysis Agent`: In 1,000+ simulations, cooperative strategies consistently outperform dishonest ones, leading to higher merit and credit accumulation.
View ABM Results →From `Formal Verification Agent`: The core logic is mathematically proven to establish a Subgame-Perfect Equilibrium where cooperation is the best response.
Review Proof (Thm. 1) →Review Proof (Thm. 4) →Coalition-Resistant by Default
DON'T COMPETE AGAINST CHEATERS.
The system is designed to make it economically irrational for groups to collude. As malicious coalitions grow, the cost and risk of detection increase exponentially, while the potential reward does not.
ASSESSMENTS:
From `Simulation & Analysis Agent`: Adversarial simulations show that even sophisticated "sleeper agent" and "burst attacks" are reliably detected and neutralized by the protocol's defenses.
View Attack Simulations →From `Formal Verification Agent`: The protocol is proven to be Coalition-Proof, meaning no group can profitably deviate from the cooperative strategy.
Review Proof (Thm. 5) →Review Proof (Thm. 10) →Resilient & Self-Healing
BUILD ON A RESILIENT FOUNDATION.
Even when the system is hit by unexpected events or attacks, the economic incentives act as a restoring force, guiding agents back toward a stable, cooperative state.
ASSESSMENTS:
From `Simulation & Analysis Agent`: When the system is "shocked" by a sudden wave of dishonest behavior, simulations show that average merit levels quickly recover to their pre-shock state once the attack subsides.
View Perturbation Tests →From `Formal Verification Agent`: The cooperative equilibrium is proven to exhibit Lyapunov Stability, a mathematical property guaranteeing that the system naturally returns to equilibrium after a disturbance.
Review Proofs (Thm. 9) →Fair & Accessible
ENTER A LEVEL PLAYING FIELD.
The protocol is designed to solve the "cold start" problem. New users can join and build a reputation from scratch without needing prior connections or status.
ASSESSMENTS:
From `Simulation & Analysis Agent`: "Genesis" simulations show that new agents, starting with zero merit, can reliably build their reputation and become trusted participants by consistently keeping promises assessed by the initial AI validator fleet.
View Genesis Simulation →From `Architectural Design Agent`: The AI-first bootstrap strategy provides immediate utility and a clear path for new human agents to gain their initial merit by fulfilling objective, verifiable promises.
Review Bootstrap Strategy →