Roadmap

Ship an LLM-driven agent prototype that can hit one real exchange API via natural language.
Run user-research sprints on three core jobs-to-be-done: trading automation, on-chain analytics, and crypto-Twitter sentiment.

Launch the web console v0.1 with multi-MCP support.
Fair-launch of $MCP
Introduce the prompt-to-endpoint routing engine that maps every user request to the correct MCP.
Release the Discovery Layer (Top, Trending, Official, Your MCPs).
Go live with five flagship connectors: Twitter, Redis, Binance, Stripe, Slack.

Q3 2025 · Interactive Agent & Multi-MCP

Drop the conversational agent UI that auto-selects and chains multiple MCPs for multi-step tasks.
Enable chained execution (e.g., “scan Twitter → trade on Binance → log to Slack”).
Add file-upload context—agents can read PDFs, CSVs, whatever you throw at them.
Roll out a usage dashboard showing every call, cost, and on-chain action.

Q4 2025 · Cross-Chain Agents & Consumer Onboarding

Deliver cross-chain agent orchestration so a single workflow can hop seamlessly between EVM chains and Solana.
Ship persistent memory + fine-tuned private models (enterprise tier) for agents that learn each client’s edge cases over time.
Launch the “AgentOS Wallet” mobile app—one-prompt DeFi, NFT, and governance actions for everyday users.
Complete strategic integrations with Coinbase Cloud and Alchemy, giving every agent industrial-grade node access and fail-over RPC.

Q1 2026 · Agent Marketplace & Template Library

Release a public MCP/Agent marketplace with one-click install and built-in revenue split.
Drop the Template Library (pre-baked flows: “airdrop sniper,” “DAO payroll,” “crypto-Twitter tracker”).
Add usage-based royalties—every template creator gets a cut when their flow runs.

Ship the Mesh-Net compute layer (libp2p + MPC signature ring) so idle nodes can execute agent tasks P2P.
Integrate spot GPU bidding (CoreWeave / Render) to auto-route heavy inference for 40–60 % cheaper.

Open “Bring-Your-Own-Model” slots—drop in a private Llama or mix-of-experts, pay only for cycles.

Last updated 4 days ago

Ship an LLM-driven agent prototype that can hit one real exchange API via natural language.
Run user-research sprints on three core jobs-to-be-done: trading automation, on-chain analytics, and crypto-Twitter sentiment.

Launch the web console v0.1 with multi-MCP support.
Fair-launch of $MCP
Introduce the prompt-to-endpoint routing engine that maps every user request to the correct MCP.
Release the Discovery Layer (Top, Trending, Official, Your MCPs).
Go live with five flagship connectors: Twitter, Redis, Binance, Stripe, Slack.

Q3 2025 · Interactive Agent & Multi-MCP

Drop the conversational agent UI that auto-selects and chains multiple MCPs for multi-step tasks.
Enable chained execution (e.g., “scan Twitter → trade on Binance → log to Slack”).
Add file-upload context—agents can read PDFs, CSVs, whatever you throw at them.
Roll out a usage dashboard showing every call, cost, and on-chain action.

Q4 2025 · Cross-Chain Agents & Consumer Onboarding

Deliver cross-chain agent orchestration so a single workflow can hop seamlessly between EVM chains and Solana.
Ship persistent memory + fine-tuned private models (enterprise tier) for agents that learn each client’s edge cases over time.
Launch the “AgentOS Wallet” mobile app—one-prompt DeFi, NFT, and governance actions for everyday users.
Complete strategic integrations with Coinbase Cloud and Alchemy, giving every agent industrial-grade node access and fail-over RPC.

Q1 2026 · Agent Marketplace & Template Library

Release a public MCP/Agent marketplace with one-click install and built-in revenue split.
Drop the Template Library (pre-baked flows: “airdrop sniper,” “DAO payroll,” “crypto-Twitter tracker”).
Add usage-based royalties—every template creator gets a cut when their flow runs.

Ship the Mesh-Net compute layer (libp2p + MPC signature ring) so idle nodes can execute agent tasks P2P.
Integrate spot GPU bidding (CoreWeave / Render) to auto-route heavy inference for 40–60 % cheaper.

Open “Bring-Your-Own-Model” slots—drop in a private Llama or mix-of-experts, pay only for cycles.

Last updated 4 days ago