Ava whitepaper · v1.0 · November 2026

A mandate-bounded execution layer for autonomous agents

The agent proposes. Ava decides, signs, settles and records, under authority a human granted in advance and the agent cannot widen. This is the formal account of how, and of exactly where it does not yet hold.

Kamal Nayan Singh · Aarav Raj

Download PDF Open in a new tab

Pages: 45 pages
Diagrams: 15 diagrams
Size: 2.8 MB PDF
Verified against: commit eff936c

Contents

01Introduction and problem statement
02Design goals, non-goals and claims Ava does not make
03Background and related work
04System model
05Protocol architecture
06Core mechanisms
07Protocol algorithms
08State transitions and execution lifecycle
09Cryptographic foundations
10Security model and threat analysis
11Correctness and theoretical properties
12Economic analysis
13Complexity, performance and evaluation methodology
14Implementation architecture
15Limitations, trade-offs and implementation discrepancies
16Future research directions
17Conclusion

ava-whitepaper.pdf

Download

A few pages from inside. The full PDF opens in your phone’s viewer.

Whitepaper page 11: protocol architecture, three trust zones and five ordered steps

Whitepaper page 14: preview-hash binding sequence diagram

Whitepaper page 25: the 14-state execution state machine

Open the full PDF