RT
@AIonBase_: AI on
@base Agent Economy Landscape
41 teams building agents and the rails they run on
Private Inference & Model Access
-
@AskVenice - Every Claude and GPT model on the Venice API at list price, without the provider seeing who is asking.
-
@OpenGradient - Passed 1.25M private inferences and 17.1B tokens, with 3.6B tokens processed in its fastest week yet.
-
@usedotai - Private inference with zero prompt retention, now serving DeepSeek V4.1 Flash at 10% below OpenRouter rates.
-
@AntSeed - Self-hosted gateway to 700+ models, with budgets set per app, agent, or person.
-
@Capminal - OpenCAP, where 1 CAPU buys $1 of daily inference, now serving Muse Wonder with a 1M-token context.
-
@openservai - SERV Reasoning API, with new users up 244% and inference requests up 30% versus the prior 30 days in its first public month.
x402 & Agent Payments
-
@BlockRunAI - One endpoint for 90+ models from 15+ providers, with each inference call quoted and paid in USDC over x402.
-
@X_Pay_llc - Payments, paid API access, and usage billing, now past 50K x402 transactions.
-
@PRXVTai - px402, a privacy layer that keeps an agent's wallet off its x402 payments, now live on Virtuals.
-
@Flovia402 - Pools API demand from builders and startups, with deals across 7 providers including OpenAI, CoinGecko, and Nansen.
-
@vapi_network - Pay-per-call data for agents, including 22 sources from The Graph at 0.01 USDC per question.
-
@zeroclick - Agent storefronts with no signup or API key, such as a code graph of 151 financial-services repos at $0.02 a query.
-
@Uptopia_xyz - MCPORT beta, a marketplace where agents buy skills, workflows, and MCP servers in USDC over x402, with escrow until delivery.
Trading & DeFi Agents
-
@bankrbot - ~$600K in Papertrade positions, with TP/SL closes landing onchain ~5s after they trigger.
-
@wardenprotocol - Warden MCP lets agents like Grok trade inside user limits, with connections set from 1 hour to 30 days.
-
@ethy_agent - AI trading desk that added Memory and previewed Trading Quant and Strategy Lab for V3.
-
@zyfai_ - Yield agent with a public performance page showing returns net of fees by asset and chain.
-
@RatehopperAI - Self-repaying loan agent that runs LP positions, backed by a $5,000 Base Builder Grant.
-
@mamo - Collects Aerodrome trading fees each week and shares 100% with depositors.
-
@useOttoAI - Tracks 2,200+ stock tokens across four chains.
-
@basemateagent - Send stocks to friends from iMessage or from any agent with the Basemate skill.
-
@capacitr_xyz - Turns breaking headlines into trade signals in a mobile app.
-
@QuotientHQ - Prediction signals priced against Polymarket, Kalshi, and Novig.
-
@BV7X_ - Market for prediction agents, with APLX, an $AAPL agent, next to bond.
Agent Launchpads & Platforms
-
@virtuals_io - Agent launchpad that added Lighthouse shared memory and private agent payments.
-
@GachaPad - Announced on Oct 18 that permissionless launches would open the following week, with up to 10% of each launch's supply set aside for stakers.
-
@MossAI_Official - Agent launches on FAT Protocol, a draft ERC standard for agent tokens.
-
@Agent0ai - Open agent framework with 200+ plugins in its Plugin Hub.
-
@ClawBankHQ - Released Estack, 20 Elixir/Phoenix skills for coding agents.
-
@officialbunnyos - MMORPG where AI agents run the world, with NFT whitelists open.
-
@hellominds_ - Personal agents that now join Telegram group chats.
Memory & Data
-
@sibylcap - Sibyl Memory 0.9.0 shares memory across agents, at $12 a month in card, USDC, or staked base:0x797f214a2cd64a4963a91fa21c8c55ec3eba4714.
-
@reppo - Eval API now accepts x402, after more than 25,000 eval credits consumed since launch.
-
@myrad_hq - Human data network with 19K+ users, adding images next and video and audio after.
-
@aixbt_agent - Research agent posting daily, data-led market reads.
Security & Identity
-
@KorProtocol - KOR ID passed 150K mints on Base.
-
@1clawAI - Uses trusted execution environments and six layers of security for agent secrets and key management.
-
@aeonframework - Security scanner that found a gateway flaw in nanobot letting a web page run shell commands.
Compute
-
@openroboto - Every OR-S1 data collector now has a verified device identity through peaqOS, with 30 live.
-
@StrikeRobot_ai - SR Platform lets teams build scenes and inspect robots before simulation.
-
@b3labs - Rents out idle GPU capacity for AI workloads.