Traceable identity
REFERENCE DEMO · v0.1.0
Make agent execution
observable.
The TITMAS Digital Cell Demo is a minimal reference implementation showing how an AI agent can carry identity, boundaries, execution evidence, and health state.
REFERENCE_COMPLETEREPLAY_PASSAPACHE-2.0Cell
WHY DIGITAL CELL
Being executable does not make an agent understandable.
Future agent systems need to answer four basic questions: who it is, what it may do, what actually happened, and what state it is in after an anomaly.
Explicit behavior boundary
Execution evidence
Health observation
DIGITAL CELL ARCHITECTURE
Six simple, bounded, composable objects
A Digital Cell does not create an unlimited autonomy center. Each object has one explicit responsibility: Evidence is not Permission, Health is not Certification, and Reputation is not Authority.
Identity
IdentityIdentify the subject, version, purpose, and source.
Boundary
BoundaryDeclare permitted behavior, prohibited behavior, and capability limits.
Evidence
EvidenceRecord intent, execution, result, and integrity summaries.
Health
HealthDerive observable state from existing records; this is not certification.
Memory
MemoryRetain execution, failure, recovery, and evolution history.
Reputation
ReputationSummarize reliability and evidence-quality history.

BOUNDED LIFECYCLE
One execution becomes traceable history.

VERIFIED REFERENCE RUN
Reproducible, without over-interpretation.
The frozen reference run passes 7 tests, produces 12 evidence records, and passes deterministic replay and SHA-256 integrity checks.

RUN LOCALLY
One implementation, one entry point, zero third-party dependencies.
Python 3 is required. The release and source are available in the public GitHub repository.
git clone https://github.com/joy7758/titmas-demo.git
cd titmas-demo/examples/digital-cell
python3 -m unittest discover -s tests -v
python3 run.pyOpen the GitHub repository →INTERPRETATION BOUNDARY
What this demo explicitly does not prove
It only shows how an agent can become observable, verifiable, and health-assessable. A health view is not a safety guarantee, and evidence integrity does not make an agent correct.