I’ve been building PRAMA-Dynamagh, a domain-agnostic control layer for humans, applications, and autonomous agents operating through Telegraph Protocol.
Autonomous agents can themselves originate mandates, acquire intelligence through Telegraph, spend delegated budget, and continue acting machine-to-machine under persistent identity, economic authority, and trajectory governance.
The basic problem is simple:
Receiving intelligence is not the same as being authorized to act on it.
PRAMA-Dynamagh turns a mandate issued by a human, an application, or an autonomous agent into an auditable execution trajectory.
Mandate
→ Agent Identity
→ Delegated Authority
→ Intent Resolution
→ Telegraph / x402
→ Miner Intelligence
→ Evidence
→ PRAMAgraph Structural Evaluation
→ Decision Gate
→ G12 Economic Authority
→ G13 Trajectory Authority
→ Autonomous Execution
→ Verifiable Ticket
→ Replay / Audit / On-chain Commitment
Human / Organization
↓
delegates authority
↓
Autonomous Agent ─────→ PRAMA-Dynamagh
│ ↓
│ Telegraph / x402
│ ↓
│ Evidence
│ ↓
│ Decision + G12 + G13
│ ↓
└──────────── continues acting
The system preserves provenance, binds decisions to evidence, supports multi-intent intelligence acquisition, and governs autonomous execution through persistent agent identity and versioned delegated authority.
A key distinction is between local decision safety and trajectory safety:
PERMIT_local + AUTHORIZE_economic + AUTHORIZE_trajectory → EXECUTE
A sequence of individually permitted decisions does not necessarily imply that the accumulated trajectory remains viable.
The current implementation includes real Telegraph miner acquisition, x402 payments, persistent evidence lineage, deterministic tickets and replay, Base Sepolia commitments, autonomous M2M execution, G12 economic governance, O_AGENT longitudinal observation, enforced G13 trajectory authority, and versioned AgentAuthorityProfile.
Repository: