Signal

Ideas shaping the agent economy.

An agent walking toward a turnstile checkpoint while a human hand lowers a price tag above it; a receipt with a check mark waits on the other side

LatestAug 2026PaymentsAgents

AG-UI Pauses the Agent. x402 Prices the Call.

CopilotKit's human-in-the-loop primitive is exactly the UX a paying agent needs before it spends. A runnable AG-UI + BlockRun walkthrough: the agent hits a 402, the price is decoded from the response, anything over a threshold waits for a human, and approval travels as shared state the model cannot forge.

Read the article

Terminal window with a curl command returning a chat completion, next to an open padlock

Aug 2026ModelsAgents

Free LLM API, No Key: Models You Can curl Right Now

Seven free, OpenAI-compatible LLM endpoints with no API key, no wallet, no signup — verified live today with a bare curl. Model list with context windows, quickstart in curl/Python/JavaScript, and the honest limits.

Illustration of a small set of sockets plugging capabilities into a terminal-based coding agent

Aug 2026Agents

The MCP Servers Worth Adding to Claude Code (2026)

Most MCP setups are bloat: servers duplicating what Claude Code does natively, burning context on tools that never get called. Here's a minimal setup that earns its keep — browser, GitHub, live docs, memory — and the one capability MCP alone can't give your agent.

Infrastructure layers of an agent-payable API stack

Aug 2026AgentsPayments

100+ Live API Endpoints, Invisible to AI Agents

Every handler served. Nothing was broken. But an agent that reads a manifest before it calls could not see 100+ live endpoints — because the manifest restated the API instead of deriving it. What an audit of three published surfaces found.

OpenClaw cost and reliability — four structural problems ClawRouter fixes via local routing

Jun 2026RoutingAgents

OpenClaw Is Burning Your Budget and Stalling Your Agents — 4 Fixes From the Issue Tracker

We read OpenClaw's issue tracker for the problems users hit in production: cost spikes from background work, provider outages that deadlock sessions, no spend ceiling for autonomous agents, one model for every task. ClawRouter's local routing, multi-provider fallback chains, and per-run cost caps fix all four.

Franklin AI agent with a Base-native USDC wallet — economic agent paying for outcomes

Apr 2026AgentsPayments

Franklin: The First Economic Agent — An Open-Source AI That Spends Its Own USDC on Base

Franklin invents a new category: the Economic Agent — software with a wallet that prices its own actions, spends toward outcomes, and stops at a hard budget cap. Open source (Apache-2.0), TypeScript, 55+ models, 16 built-in tools, USDC settlement on Base & Solana via x402. YOPO pricing — you pay only for outcomes.

Franklin — the AI agent with a wallet, spending USDC per action via x402

Apr 2026Agents

Franklin: The AI Agent With a Wallet — Pay-Per-Action Autonomous Coding

Franklin is the first Autonomous Economic Agent — it spends USDC per action under a wallet-enforced budget cap. 55+ models, 12 built-in tools, free to start on 9 NVIDIA-hosted models (DeepSeek V4, Qwen3, Nemotron Nano Omni vision, and more).

Jan 2026Agents

X/Twitter Algorithm: How AI Agents Can Hack Organic Reach

Reverse engineering X/Twitter's algorithm for agent-native content. Learn how AI agents can work with the algorithm to maximize organic reach and virality.