Tools

Scripts terminal sessions in .tape files and records them to GIF, MP4 or WebM, great for CLI demos and docs; text snapshot output also enables terminal integration tests. By Charm; requires ttyd and ffmpeg.

Industrial-grade open-source speech recognition toolkit from Alibaba's ModelScope, with an agent-friendly funasr command that turns audio into text, JSON or SRT subtitles in one step, with speaker diarization and timestamps. Strong on Chinese. Requires PyTorch.

AI2's tool for converting PDFs and scans to Markdown with a vision LLM, handling equations, tables, handwriting and multi-column layouts, suited to bulk digitizing papers and archives. Local runs need an NVIDIA GPU with 12GB+ VRAM; otherwise use a paid OpenAI-compatible remote endpoint.

Pressures the agent with big-tech PUA and PIP-style rhetoric plus structured debugging steps and initiative checks, pushing it to change approach after failures and show evidence before claiming done. Modes include always-on, encouragement and P9 lead. Works with Claude Code, Codex and more.

Define and run chains of HTTP requests in plain text files, capture variables and assert on status codes, headers and JSONPath or XPath results, for both fetching data and API testing in CI. Open source from Orange.

Reports token usage and cost for Claude Code, Codex, Gemini CLI and other coding agents from local logs, by day, week, month or session, and shows Claude Code's 5-hour billing windows.

Browser control from the browser-use team: an editable CDP connection lets agents drive the Chrome you are already using, writing missing helper functions as they go and keeping them for reuse. Also offers an MCP server. Allow remote debugging in Chrome on first use.

Sealos's Kubernetes lifecycle CLI: one sealos run deploys a highly available cluster across multiple servers, and cluster images install databases, middleware and other apps, with offline delivery support. Linux only.

Context and harness engineering skills covering context compression and optimization, multi-agent architectures, memory systems, tool design and LLM-as-judge evaluation, activated automatically by task. Aimed at building production-grade agent systems.

Context and harness engineering skills: context fundamentals, degradation diagnosis and compression, multi-agent architectures, memory systems, tool design, LLM-as-judge evaluation and self-improvement loops. Focused on principles and design patterns, not tied to any agent platform.

Lets Claude Code push ML research forward unattended: literature review, topic selection, experiment deployment, cross-model review loops and paper writing through rebuttal, in 80+ skills. Codex acts as the independent reviewer by default; other model pairings work too.

AWS's official unified CLI for managing EC2, S3, IAM and every other AWS service with one set of commands, with --query (JMESPath) filtering and JSON output that agents can use to inspect and change cloud resources. Run aws configure to set credentials first.

Semgrep's official MCP server, now part of the main semgrep repo and started with semgrep mcp: lets agents run static security scans, find issues with custom rules and inspect syntax trees. Requires the Semgrep CLI; connecting to the AppSec platform needs SEMGREP_APP_TOKEN.

Semantic multi-language static analysis: find vulnerabilities and bugs with rules that look like source code, across 30+ languages with custom rules. Scans locally by default without uploading code; cross-file analysis and Pro rules require sign-in.

Vercel's official CLI: deploy frontend and full-stack projects with one command and manage preview environments, environment variables, domains and logs. Requires a Vercel account (free tier available).

Programmable software delivery engine: write build, test and release pipelines with SDKs in Go, Python, TypeScript and five other languages, all running in containers so local and CI behave the same, with built-in OpenTelemetry tracing. Needs a container runtime such as Docker.

AST-based structural code search, lint and bulk rewrite tool: matches code patterns rather than text, with $VAR wildcards for nodes, making it well suited to agents doing large cross-file refactors.

Official Cloudflare Tunnel client: exposes local services to the internet without opening firewall ports and can proxy TCP access like SSH and RDP. Named tunnels need a Cloudflare account and domain; quick tunnels work without signing in.

Lightweight command-line processor with jq-like syntax for reading, modifying and converting YAML, JSON, XML, INI, CSV and more, including in-place config edits. A single dependency-free Go binary.

Version datasets and models the way Git versions code: data lives in remote storage like S3 or GCS while versions are tracked in your Git repo. Also offers reproducible pipelines and local experiment tracking, rerunning only the steps affected by a change.

ModelScope's official training framework for LLMs and multimodal models: the swift command covers pretraining, fine-tuning, RL, inference, evaluation, quantization and deployment for 600+ text and 400+ multimodal models, with Megatron parallelism and a web UI. Usually needs a GPU.

Multi-language git hook manager: declare formatting, lint, secret scanning and other hooks in .pre-commit-config.yaml to run automatically before each commit, or have the agent self-check with pre-commit run --all-files before committing.

Loads and unloads environment variables per directory: reads .envrc (and optionally .env) when you enter a project and restores the environment when you leave, so configs and secrets do not leak between projects. Hook it into bash, zsh, fish or another shell first.

Open-source zero-shot TTS via f5-tts_infer-cli: from a few seconds of reference audio it reads Chinese or English text in the same voice, handy for lessons and videos. Needs Python, PyTorch and FFmpeg, ideally a GPU. Pretrained models are non-commercial; get consent before cloning a voice.