ABIS — Agent Business Interaction Standard

AGENT BUSINESS INTERACTION STANDARD

BeyondZero Click.

ABIS is a framework that connectswhat you ask AI to dowith what actually happens.

Reservation assistant

AI AGENT

ILLUSTRATIVE SCENARIO

YOU

Reserve a restaurant for 4 tomorrow at 19:00.

AI AGENT

I'll find a place that fits.

WORKING
  1. SEARCH
  2. AVAILABILITY
  3. RESERVE
AI AGENT

Your reservation is set.

Reservation details19:004 guests

SUCCESSDONE

OUTCOME

SUCCESS ≠ OUTCOME CONFIRMED

INTENDED OUTCOME19:004 guests
OBSERVED OUTCOME19:304 guests

MISMATCHTIME +30 MIN

ABIS
  1. 01OBSERVE
  2. 02COMPARE
  3. 03EVIDENCE

The Internet
Was Built
Around Clicks.

For decades, doing business online has meant operating interfaces.

Finding pages. Comparing options. Filling forms. Moving between systems. Completing transactions one screen at a time.

  1. 01Search↗
  2. 02Click↗
  3. 03Navigate↗
  4. 04Compare↗
  5. 05Fill↗
  6. 06Book↗
  7. 07Pay↗
  8. 08Confirm↗

AI Agents change that assumption.

From Human Operation
to Human Intent.

THE INTERFACE ERA

Human / Interface /
Website / App / Business system

THE INTENT ERA

Human Intent / AI Agent /
Business Interaction / Business Outcome

Humans no longer need to operate every interface themselves. They can express intent and delegate execution to AI agents.

But removing the click
does not remove the
business interaction.

Technical Success
≠Business Success.

Technical success does not guarantee business success. HTTP 200 does not mean user intent was achieved.

An AI agent can execute a system successfully and still fail to reach the business outcome the user intended. ABIS is concerned with what actually happened in the business interaction—not merely whether execution completed.

Tap to enlarge

A system can execute successfully while the intended business result is not achieved.

AI agents need more than execution status. They need a common way to understand what happened in the business interaction and what outcome was actually achieved.

ABIS — Agent Business Interaction Standard

A COMMON WAY TO UNDERSTAND

Across agents.
Across protocols.
Across businesses.

ABIS is a framework designed to support interoperability by providing a common way to represent Business Interactions and Business Outcomes across different AI agents, protocols, and business systems. Interoperability between independent third-party implementations has not yet been established.

AI agentsExisting protocols / executionBusinessBusiness InteractionBusiness Outcome

A conceptual relationship, not a required middleware architecture.

Business
Interaction

What happened between the agent and the business?

ABIS treats the business interaction itself as an explicit shared concern across agents, protocols, and business systems.

01

Reservation

02

Purchase

03

Application

TECHNICAL EXECUTION✓
≠
BUSINESS OUTCOME?

Did the intended business
result actually happen?

A successful API call is not the same thing as achieving the result the user intended.

One Purpose.
Multiple Business
Interactions.

A single human purpose may span multiple businesses, systems, and interactions.

ABIS explores how those interactions and their outcomes can be understood across system boundaries.

One human purpose can span multiple businesses, systems, and interactions—from travel and lodging to payments, dining, purchases, and administrative steps—before a single business outcome can be evaluated.

Tap to enlarge

Company and service names are examples only. They do not indicate partnership or ABIS compatibility.

ABIS DOES NOT REPLACE AGENT PROTOCOLS.

Protocol Neutral.

AI AGENTS
MCPA2AUCPAPIsOther systems

Business Interactions / Business Outcomes

Different environments.
A shared business concern.

Existing protocols help agents connect, communicate, discover capabilities, and execute actions.

ABIS focuses on the business interaction and business outcome across those environments.

ABIS is complementary.

Public v0.1

01

Concept & Architecture Release

PUBLICNON-NORMATIVEOPEN FOR REVIEW

ABIS v0.1 introduces the public concept, terminology, scope, architecture, examples, governance, and roadmap.

It is an early public release intended for discussion, validation, and ecosystem feedback.

Specification

ABIS Public v0.1 concept, terminology, scope, architecture, examples, governance, and roadmap.

Explore the Specification

Interactive Demonstrator

A non-normative demo across five business cases showing Business Interaction and Business Outcome.

Try the Demonstrator
DEVELOPER PREVIEWv0.8.0

ABIS Reference Runtime v0.8.0

Runnable reference implementation with a machine-readable Runtime Profile, Interaction Preflight, and provider-neutral Reference Agent Client. Validate the Profile → Preflight → Invoke flow from a known Runtime URL.

Run the Reference Runtime

From Concept
to Adoption.

  1. 01

    Public concept

  2. 02

    Industry validation

  3. 03

    Multi-company PoC

  4. 04

    Interoperability validation

  5. 05

    Specification evolution

  6. 06

    De facto adoption (future aspiration)

LONG-TERM ASPIRATION (NOT YET ACHIEVED)

Global Validation → International Standardization

An intended direction and future aspiration — not a current achievement or a guaranteed outcome.

Frequently Asked Questions

Key facts about ABIS, Public v0.1, and the Reference Runtime — structured for search and AI answer engines.

ABISABIS (Agent Business Interaction Standard) is a framework designed to support interoperability by representing Business Interactions and Business Outcomes across AI agents, protocols, and business systems. Interoperability between independent third-party implementations has not been established.

What is ABIS?
ABIS (Agent Business Interaction Standard) is a framework designed to support interoperability by representing Business Interactions and Business Outcomes across AI agents, protocols, and business systems. Interoperability between independent third-party implementations has not been established.
What is ABIS Public v0.1?
ABIS Public v0.1 is the Concept & Architecture Release — the first public release of concept, terminology, scope, architecture, examples, governance, and roadmap. It is not a finished standard. It is intended for discussion, validation, real-world evaluation, and ecosystem feedback.
What is ABIS Reference Runtime v0.8.0?
It is a publicly released reference implementation with Reference Runtime Profile, Interaction Preflight, and Reference Agent Client. From a known Runtime base URL, you can verify Profile → Preflight → Invoke. Public Validation is active. REAL_EXECUTION is prohibited. Conformance, certification, and Internet-wide discovery are not implemented.
Does ABIS replace MCP, A2A, or other protocols?
No. ABIS does not replace existing protocols. Protocols provide connection, messaging, capability discovery, and action execution. ABIS focuses on representing Business Interactions and Business Outcomes across those environments.
What is the difference between Technical Success and Business Success?
A successful HTTP response or system execution does not mean the user's intended Business Outcome was achieved. ABIS addresses how to understand what happened in a Business Interaction and which Business Outcome was reached — beyond execution status alone.
Does ABIS provide certification or conformance today?
In the public Reference Runtime v0.8.0, conformance, certification, registry, and Internet-wide discovery (/.well-known/abis) are not implemented. Public materials do not claim ABIS Compatible, Conformant, or Certified status.
Who operates the ABIS project?
CoaRetail G.K. leads the ABIS project. Official site: https://abis.coaretail.com. Public specification: https://github.com/abis-standard/abis.

10A PUBLIC VALIDATION

Published, under validation —
evolving with evidence.

ABIS Reference Runtime v0.6.0 was independently validated from different AI agent environments, using only the public repository as the source of truth.

Friction found in the initial validations informed the public documentation. New sessions from the same provider cohort then revalidated all three targeted improvements.

REFERENCE RUNTIME v0.6.0Observations within this validation cohort.
VALIDATION RUNS
6

3 initial + 3 revalidation

AI AGENT PROVIDERS
3

Different AI agent environments

TARGETED FIXES
REVALIDATED
3 / 3

Targeted improvements

OBSERVED SEMANTIC
BOUNDARY FAILURES
0

within this validation cohort

EVIDENCE-DRIVEN IMPROVEMENT

Turn validation records
into the next improvement.

  1. 01VALIDATION

    Validate from public information

  2. 02EVIDENCE

    Record observations and constraints

  3. 03FEEDBACK

    Reflect in public documentation

  4. 04REVALIDATION

    Confirm in a new session

↶Toward the next validation — a continuing improvement loop
INITIAL EVIDENCEEV-001 / EV-002 / EV-003
DOCUMENTATION FEEDBACKVFP-01
REVALIDATION EVIDENCEEV-004 / EV-005 / EV-006
EVIDENCE REVIEWCAVER-01

TARGETED FIXES / REVALIDATION

VALIDATED indicates improvement of the targeted documentation or validation friction.

01ChatGPT / OPENAI
INITIAL VALIDATIONEV-001

Environment /
Validation-Mode Gap

VFP-01Public Documentation
REVALIDATIONEV-004

Targeted friction improvement validated.

VALIDATED
02Claude / ANTHROPIC
INITIAL VALIDATIONEV-002

Observe
Discoverability Gap

VFP-01Public Documentation
REVALIDATIONEV-005

Targeted friction improvement validated.

VALIDATED
03Grok / xAI
INITIAL VALIDATIONEV-003

Version
Selection Gap

VFP-01Public Documentation
REVALIDATIONEV-006

Targeted friction improvement validated.

VALIDATED

Independent validation evidence.
Not certification or conformance determination.

These observations do not imply endorsement, approval, or certification by any AI provider and are limited to this validation cohort.

JOINT VALIDATION

Verifying that an AI agent's work is actually done — across companies.

ABIS is pursuing joint validation and technical dialogue with companies and research institutions on how to confirm the business results of AI agents.

ABIS Joint Validation & Technical Dialogue

Verifying that an AI agent's work is actually done — across companies.

AI agents are beginning to carry out work — reservations, purchases, payments, applications — across multiple companies and services.

Yet even when each individual system works correctly, that does not guarantee that the work the user asked for was completed as intended.

ABIS (Agent Business Interaction Standard) is a public technical specification for confirming the business results of AI agents by linking what was requested, what was executed, the supporting evidence, and the completion conditions.

We currently publish a public specification candidate, a reference implementation, and validation results, and are advancing technical validation while stating their limitations explicitly.

We welcome joint validation and technical discussion with companies and research institutions.

If you are interested in confirming business results across different AI agents, companies, and services, please contact us.

Note The ABIS public specification candidate has validation limitations. Interoperability between third-party implementations and formal conformance certification have not been established at this time. Rights conditions for implementing or reusing the specification must also be confirmed separately.

Contact us about joint validation (Opens in a new tab)

JOIN THE VALIDATION

JOIN THE
VALIDATION

Publication is only the beginning.

Together with agent builders and organizations that operate business systems, we will explore interoperability.

FOR AGENT BUILDERS

Validate ABIS
with your agent.

Independently validate ABIS Reference Runtime using only the public repository as the source of truth.

Findings, ambiguities, and operations you could not execute all contribute to validation evidence.

Repository→Profile→Preflight→Invoke→Evidence

Operations you could not execute still record their constraints.

VALIDATE WITH YOUR AGENT (Opens in a new tab)

GitHub / Agent Validation Prompt

FOR BUSINESS / PLATFORM PARTNERS

Explore interoperability
with your business system.

Across reservations, commerce, SaaS, enterprise systems, and public services, jointly explore whether Business Interaction between agents and business systems can be handled consistently across different agent environments.

AI AGENT AAI AGENT BAI AGENT C
ABISBusiness Interaction / Business Outcome
explored as a common framework
BUSINESS SYSTEM

Protocol and execution remain external mechanisms.

START A VALIDATION (Opens in a new tab)

Contact us about joint validation

VALIDATE TOGETHER

Validate. Record. Improve.
Create the next evidence together.

Beyond Zero Click.

From technical execution to business outcome.