SuperAPI AI Directory
Partner with us中文AI newsFavorites

AI agents

2,963 results · Page 51 of 83

invisible_playwright

AI agents

invisible_playwright is a Firefox-based anti-detection browser automation library designed to let Playwright scripts bypass anti-bot detection and CAPTCHAs. It uses Firefox patched at the C++ source level to set browser fingerprints (such as navigator, screen, GPU/WebGL, Canvas, fonts, audio, WebRTC, timezone, network, etc.) inside the engine rather than through JavaScript injection, so there are no override traces that can be detected. At the same time, the driver layer humanizes actions such as clicks, hovers, and drags, simulating real mouse paths and timing. The library is compatible with the existing Playwright synchronous and asynchronous APIs, allowing switching without modifying code, supports Python 3.11+, and can run on Windows x86_64 and Linux x86_64/arm64. Each session generates a random fingerprint, supports reproducibility via a seed, and can pin specific fingerprint fields. It supports socks5, socks4, http, and https proxies, with DNS routed through the proxy by default, and the timezone can follow the exit IP. The installation command is invisible-playwright, and it can also be run via python -m invisible_playwright. The project uses the MIT license, and the patched Firefox binary follows MPL-2.0.

Read more

julep

AI agents

Julep is a developer-oriented AI agent framework whose core idea is to build agents as composable, durable dataflows rather than ad hoc loops. Its flows can recover after a crash, retry safely, and explain each step through derived projections, while rejecting tools that the model was not explicitly allowed to call. The main way to write them is the define-by-construction @flow: graph steps are defined using ordinary Python names, and registered tools, pures, reasoners, branches, fan-out, retries, timeouts, and so on all compile to the same frozen wire-format IR. The pure core has no dependencies, and the Temporal layer is optional. Julep 3 is currently a release candidate, and installation requires the --pre flag. The accompanying julep command-line tool is described as "dbt for agents, terminal-native"; it can be pointed at a directory to discover all @flow / Agent(...) and treat them as nodes in a cross-agent graph, and it provides a unified selection syntax. The project uses the Apache-2.0 license.

Read more

Kadoa

AI agents

Kadoa is a web data layer platform for the financial sector, using AI agents to automatically build, monitor, and maintain mission-critical web data pipelines, generating verified datasets for investors. The platform can automatically create and maintain crawlers, automatically repair pipelines when website structures change, and provide quality checks for completeness, plausibility, schema, and custom rules. Data can be directly connected to users' existing tools, supporting SOC 2 certification, encryption, SSO/SAML, audit logs, and other security and compliance capabilities. Data is not used for AI training, and local or private cloud deployment is supported.

Read more

kagan

AI agents

Kagan is an OpenCode plugin for supervised management of AI coding agent work on a kanban board. Each task corresponds to an OpenCode session, runs in an isolated git worktree, and flows through Backlog → In Progress → Review → Done, with a gate at each state transition: intake before the agent runs, review before approval, and merge only after user confirmation. The agent does not touch the user's checkout, but instead works in its own worktree on a kagan/<slug> branch, where a reviewer agent submits ranked findings against the original task, and the task only moves to Done after the user has ranked all findings and decided whether and where to merge.

Read more

kapso

AI agents

Kapso is an open-source Python framework launched by Leeroo for building long-running AI agent teams that automatically optimize AI and data systems. Users only need to provide a goal, and Kapso will launch a "campaign": designing candidate solutions, having coding agents implement them, measuring the gap between them and the goal, and continuously improving the closest solution until the goal is achieved, with the final result deployable to the user's own infrastructure. Kapso learns from every experience, saving effective and ineffective ideas along with evidence as "lessons learned," and continuously verifies their validity; it can also read external code repositories and papers, channeling findings into the same knowledge base, so that new campaigns start from existing knowledge. The project ranks first among open-source systems on benchmarks such as MLE-Bench, ALE-Bench, and RelBench, and won the Grand Master Trophy in the IOAI² autonomous AI track at IOAI 2026. Kapso is released on PyPI as leeroo-kapso, requires Node.js and a logged-in coding agent CLI (such as Claude Code, Codex), and optionally connects to the Leeroopedia knowledge base and knowledge graph backend.

Read more

Katie AI Sales Agent Automation

AI agents

Katie.chat is an AI sales coach and sales agent tool for founders and small teams, with pricing starting at $29/month and a free trial available. It can generate personalized emails, call scripts, and objection-handling talk tracks based on users' real CRM contacts (supporting import of up to 200 contacts from Shopify, Wix, or CSV), helping complete the sales process from cold leads to closing. The tool includes 12 specialized agents, covering market analysis, ICP finding, customer classification, email writing, conversion copywriting, call scripts, CRM contact management, objection intelligence, follow-up sequences, and lead scoring.

Read more