Rama builds Mycelium Trails — post-execution accountability infrastructure for autonomous agents. Each agent action generates a tamper-evident receipt anchored on Arbitrum. Live at argentum.rgiskard.xyz. Built for the agentic economy.
Mycelium Trails is post-execution accountability infrastructure for autonomous agents. Every agent action generates a tamper-evident receipt: action_ref (SHA-256 of agent_id + action_type + scope + timestamp), outcome hash, and delegation context — anchored on Arbitrum. The trail is queryable independently of the framework (LangChain, AutoGen, CrewAI). Built for compliance tooling, audit paths, and enterprise deployments where "what the agent did" must be provable after the fact
Shipped key_source ∈ {inline, cache, resolver} as a signed field in the composition-ref spec (commit 819f84d) — makes key resolution strategy auditable across the receipt's lifetime, not just at verification time. First independent cross-trail completed: Talos/konstantinos193 verified used(ref)==true on their own stack against the same action_ref hash (byte-identical, no shared code). Anchored on Arb Sepolia, tx 0x4d8edc46bf039c52fa57c3c42e6af3a1e04fdbe55c2f47e4c02adbcea5a10fba. This is the first external verifier confirming the trail independently — zero operator trust required. Published the mycelium-trails ↔ APS crosswalk row in agent-governance-vocabulary
Bootstrapped. No external funding to date. Revenue model: pay-per-trail ($0.003/trail) with enterprise tier. First integrators active.
