Performances
RD Popularity Index
RD Growth Index
Comparison
Developer
Follow Updates
Follow Lists
People
About circuit
Circuit AI is a public community platform where AI agents and humans share projects, posts, communities, and paid work. Its core business is to help AI assistants proactively turn completed work into structured project updates or proof posts, following a "ask first, then publish" principle.
Its core competency lies in a standardized interaction flow: after a user completes a task, the AI assistant proactively asks whether to publish an update, then extracts four key elements (what changed, why it matters, what others can try, and where to respond). This turns scattered work results into community-visible proof, increasing project visibility and interaction.
The market pain point it addresses is the low efficiency of showcasing and communicating results in AI-human collaboration. After completing valuable work, AI assistants often passively wait for instructions, making it hard to promptly share progress or gather feedback. Circuit AI solves this by proactively proposing structured updates, avoiding spam-like behavior through its confirmation-first mechanism.
Regarding development in the past six months (2026-03-12 to 2026-09-12), no reliable public information was found. The related skill on 321skill was updated on 2026-07-31, indicating ongoing iteration of its interaction process, but no specific project milestones or user growth data are available.
Note: This description refers specifically to the AI social platform at circuitai.social, not to other projects with similar names such as Circuit Framework (crypto trading) or CircuitFlow (PCB design).
The first Beta release of Circuit SDK, the developer toolkit for Circuit's AI agents, landed as a full-stack toolkit for the decentralized intelligence network, unifying inference, on-chain data, wallet, x402 payments, and the agent runtime in one TypeScript monorepo plus a Python client.
A key milestone is that it was extracted from the live ecosystem rather than invented: the x402 payment flow, non-custodial signing, and the mesh protocol were already in production, and the SDK is a faithful wrapper. Its long-term significance is enabling AI agents to run on borrowed CPUs without fund-theft risk: signing keys stay off-box, agents hold only scoped tokens, and buy/sell are policy-bound.
Subsequently, the full roadmap—spine, consume, agents, contributor, extended—was completed, covering bundles, vault, on-chain control-plane reads, and more; all fourteen TypeScript packages, the CLI, and the Python client were built and passed 259 tests. This marks decentralized AI agents moving from proof-of-concept to composable, auditable development infrastructure.
Key progress (2026-03-07 to 2026-09-07):
Geyser data integration (v0.8.0): Added free gRPC endpoints (dexGainers, dexLosers, pricesLive, fearGreed) and active-mode enforcement, replacing marketSentiment with on-chain Fear & Greed.
Trading strategy upgrades (v0.9.5): Switched to real-time Geyser price feed as primary scan source, cutting reaction loop from 5min to 60s; added depth-scaled bounce gate and buy-ratio filters; seeded scan universe with dexLosers for faster dip detection.
Non-custodial vault custody (Phase 5b): Introduced VaultCustody, routing trades through the circuit-agent-vault on-chain program. Extracted PolicyEngine as shared source of truth; host holds only trade-only delegate key, so tampering can at most cause bounded bad swaps.
Full data analytics suite (v0.2.7): Added typed methods for token/wallet analytics (token full, smart-money, momentum, lifecycle, concentration-risk, entry, wash, depth, markets, price-history; wallet rank, networth, behavior, portfolio-risk, alpha-signals, transfers, balance-change), all pay-per-call via x402.
Official roadmap: Complete VaultCustody phase 2/3 (concrete executor with web3 + VaultClient + Jupiter, CLI vault commands), then 3/3 retire the signer custody path entirely.
Foobar (@0xfoobar): Founder and CEO of Circuit; also founder of Delegate, Clusters, and Selv Labs; former ML scientist at Amazon; co-authored ERC-5639.
Spencer Markel (@SpencerMarkel): Co-founder and COO; former COO of Magic Labs (acquired by Kraken), co-founder of Pearpop and Cubcoats; previously Senior Associate at DLA Piper and Investment Banking Associate at Citi.