REGLO Agent
An on-chain AI agent driven by large model decisions that schedules computing tasks to avoid solar curtailment windows and earns token rewards through on-chain proofs, implementing an architecture that separates LLM selection from cryptographic verification.
Timeline (last 90 days)
The full project address of the REGLO Agent is github.com/edmundjeevan/reglo-agent-pipeline.
REGLO Agent is an AI Agent built for the Monad Metropolis hackathon that automatically schedules compute tasks to avoid solar curtailment windows and receives MON token rewards after proving cost savings on-chain.
REGLO Agent's decision logic is centralized in the agent/reason.js file, calling Qwen through DashScope's OpenAI-compatible chat-completions interface, and setting REASONING_PROVIDER=qwen switches all decisions to Qwen.
In one real run, Qwen explained its choice: 'The 13:00 slot has the lowest electricity price, and all sites have the same curtailment risk, so this is the cheapest available option.'
The input Qwen receives includes the full list of candidate windows (site, hour, curtailment risk, electricity price), and it returns a selection result and a natural-language explanation.
On each run, REGLO Agent independently computes the deterministic cost-optimal window and compares it with Qwen's choice; if Qwen's answer is not the cheapest option, the Agent records it rather than silently trusting the model.
In all real runs described in the article so far, Qwen's choice has matched the true optimal solution.
In the REGLO Agent architecture, Qwen never touches the numbers that determine payment, only choosing which window to act in; the oracle independently re-derives the baseline and actual cost from REGLO real data and signs them; the smart contract checks the savings amount itself before payment.
The REGLO Agent is designed so that errors by the large model do not lead to loss of funds, only to a discount in the savings achieved.
The same code path of the REGLO Agent also supports Kimi and a local Ollama fallback, so that even if a certain provider is unreachable, the Agent can continue working.
10 more timeline events
All Facts (20)
The full project address of the REGLO Agent is github.com/edmundjeevan/reglo-agent-pipeline.
50%UnverifiedREGLO Agent is an AI Agent built for the Monad Metropolis hackathon that automatically schedules compute tasks to avoid solar curtailment windows and receives MON token rewards after proving cost savings on-chain.
50%UnverifiedREGLO Agent's decision logic is centralized in the agent/reason.js file, calling Qwen through DashScope's OpenAI-compatible chat-completions interface, and setting REASONING_PROVIDER=qwen switches all decisions to Qwen.
50%UnverifiedThe input Qwen receives includes the full list of candidate windows (site, hour, curtailment risk, electricity price), and it returns a selection result and a natural-language explanation.
50%UnverifiedOn each run, REGLO Agent independently computes the deterministic cost-optimal window and compares it with Qwen's choice; if Qwen's answer is not the cheapest option, the Agent records it rather than silently trusting the model.
50%UnverifiedIn all real runs described in the article so far, Qwen's choice has matched the true optimal solution.
50%UnverifiedIn the REGLO Agent architecture, Qwen never touches the numbers that determine payment, only choosing which window to act in; the oracle independently re-derives the baseline and actual cost from REGLO real data and signs them; the smart contract checks the savings amount itself before payment.
50%UnverifiedThe REGLO Agent is designed so that errors by the large model do not lead to loss of funds, only to a discount in the savings achieved.
50%UnverifiedThe same code path of the REGLO Agent also supports Kimi and a local Ollama fallback, so that even if a certain provider is unreachable, the Agent can continue working.
50%UnverifiedThe same code path also supports Qwen and local Ollama as fallbacks, so the agent can keep working even if one provider is unreachable.
50%UnverifiedThe oracle independently re-derives the baseline and actual costs from real REGLO data and signs them, and the smart contract checks the savings amount itself before paying.
50%UnverifiedThe full project address is github.com/edmundjeevan/reglo-agent-pipeline.
50%UnverifiedREGLO Agent saves costs by scheduling compute tasks outside solar curtailment windows, and automatically receives MON token payment after proving the cost savings on-chain.
50%UnverifiedIn REGLO Agent, after setting REASONING_PROVIDER=kimi, all decisions in that run are routed to Kimi.
50%UnverifiedKimi receives the full list of candidate windows (including site, hour, curtailment risk, and electricity price), then returns a choice along with a plain-language rationale.
50%UnverifiedOn each run, REGLO Agent independently computes the deterministic cost-optimal window and compares it with Kimi's choice; if Kimi's answer is not the cheapest option, the agent explicitly points this out in the log.
50%UnverifiedIn all real runs so far, Kimi's choice has matched the true optimal solution.
50%UnverifiedIn the REGLO Agent architecture, Kimi never touches the numbers that determine payment; it is only responsible for choosing which window to execute the task in.
50%UnverifiedIn one real run, Qwen explained its choice: 'The 13:00 slot has the lowest electricity price, and all sites have the same curtailment risk, so this is the cheapest available option.'
50%UnverifiedIn one run against real REGLO data, reasoning through Kimi produced a real payment on the Monad testnet, with transaction hash 0xcc5f99ebfccb2a9128a102984f5ddde4d22d67c9b4268bc9b42201e058b8e438.
50%