Skip to content

docs(paper): retitle to Agent Native Format #23

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
shreyanshjain7174 merged 1 commit into from
Jul 16, 2026
Merged

docs(paper): retitle to Agent Native Format #23

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 5 additions & 4 deletions paper/acp.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
# ACP: Intent-Resolved Execution Contracts for Autonomous Agent Tool Use
# Agent Native Format: Token-Efficient Execution Contracts for Autonomous Agent Tool Use

**Shreyansh Sancheti**
*Clawdlinux / NineVigil*
Expand All @@ -11,9 +11,10 @@
Production agent systems built on the Model Context Protocol (MCP)
spend a large fraction of their context window on tool discovery and
verbose JSON-Schema descriptors before performing any user task.
We introduce the **Agent Contract Protocol (ACP)**, a server-side layer
that sits on top of MCP and returns a single intent-scoped Execution
Manifest in one round trip with auth pre-injected, dependency order
We present the execution-contract layer of **Agent Native Format (ANF)**.
Named the Agent Contract Protocol (ACP), it is a server-side layer that
sits on top of MCP and returns a single intent-scoped Execution Manifest
Comment on lines +14 to +16
in one round trip with auth pre-injected, dependency order
pre-computed, and security boundaries declared.

We evaluate ACP across four frontier models — Claude Opus 4.7, Claude
Expand Down
11 changes: 6 additions & 5 deletions paper/acp.tex
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,10 +20,10 @@
showstringspaces=false,
}

\title{ACP: Intent-Resolved Execution Contracts for Autonomous Agent Tool Use}
\title{Agent Native Format: Token-Efficient Execution Contracts for Autonomous Agent Tool Use}
\author{Shreyansh Sancheti\\
\texttt{007ssancheti@gmail.com}\\
Clawdlinux / NineVigil}
\texttt{github.com/Clawdlinux}\\
Clawdlinux}
Comment on lines 24 to +26
\date{May 2026}

\begin{document}
Expand All @@ -32,8 +32,9 @@
\begin{abstract}
Production agent systems built on the Model Context Protocol (MCP) spend a
large fraction of their context window on tool discovery and verbose
JSON-Schema descriptors before performing any user task. We introduce the
\textbf{Agent Contract Protocol (ACP)}, a server-side layer that sits on top
JSON-Schema descriptors before performing any user task. We present the
execution-contract layer of \textbf{Agent Native Format (ANF)}. Named the
Agent Contract Protocol (ACP), it is a server-side layer that sits on top
Comment on lines +35 to +37
of MCP and returns a single intent-scoped Execution Manifest in one round
trip with auth pre-injected, dependency order pre-computed, and security
boundaries declared. We evaluate ACP across four frontier models --- Claude
Expand Down
Loading