The SwarmPane journal
Field notes for people who build with AI.
Practical ideas for agent workflows, voice, and connected tools.
Claude Fable 5.2 in Claude Code: price and setup
Claude Fable 5.2 launched on November 9, 2026. How to select it in Claude Code, what it costs on Pro, Max and the API, the launch offer, and when to use it.
Conductor is shutting down: dates and alternatives
Conductor stopped new downloads on November 4, 2026. Its Cloud shuts down November 20 and its desktop app winds down in early 2027. Here is where to move.
Claude Code GUI for Windows and Linux
Yes: Anthropic’s desktop app runs Claude Code on Windows and, in beta, on Ubuntu and Debian. Here are the other GUIs for both, checked October 2026.
Best tools to run AI coding agents in parallel
The best tools to run several AI coding agents in parallel in 2026, from their own docs: Mac, Windows and Linux support, which agents, worktrees and price.
Devin CLI: install, Windows, vs Claude Code
Install Devin CLI on macOS, Linux or Windows, sign in with devin auth login, see what Devin costs, and compare it with Claude Code, from Cognition’s own docs.
Hermes Agent: desktop app and install guide
Hermes Agent has a desktop app on Mac and Windows and a one-line install on Linux. How to set it up with Nous Portal or another provider, from its own docs.
SwarmPane vs Conductor: which fits you?
SwarmPane vs Conductor from both apps’ own pages: platforms, agents, worktrees, review and price, and what Conductor joining Meta changes for you.
Conductor alternatives for Mac, Windows and Linux
Conductor stopped new downloads on November 4, 2026 and never shipped for Windows or Linux. Alternatives for every platform, checked against their own docs.
Codex usage limit reached: what to do
Codex usage limit reached? What the message means, where to check your usage and reset time, what Plus and Pro get, and how to keep working with another agent.
Run Claude Code and Codex at the same time
Yes, you can run Claude Code and Codex at the same time. Use separate worktrees, one shared AGENTS.md, split roles and a plan for when a usage limit hits.
Cursor vs Claude Code: which should you use?
Cursor vs Claude Code compared from each vendor’s own docs: editor or terminal agent, which models you get, pricing as of October 2026, and how to use both.
Claude Code alternatives: 8 coding agents compared
Eight Claude Code alternatives checked against their own docs in October 2026: which are open source, free or work with a local LLM, and what each one costs.
Claude Max vs Pro for Claude Code: is it worth it?
Claude Max vs Pro for Claude Code: what Max adds, how to measure your own usage with /usage first, and when Max beats Pro plus pay-as-you-go usage.
How to run Claude Code on Windows
How to run Claude Code on Windows 11 or 10: the PowerShell install command, whether Git Bash is still required, WSL or native, and fixes for common errors.
Gemini CLI limits: free tier, daily caps, resets
Gemini CLI limits as of October 2026: what happened to the free tier, the daily request caps by sign-in method, when they reset, and what to do at the limit.
Claude Code /compact vs /clear: manage context
Claude Code /compact vs /clear: what each does to your context window, what survives compaction, how auto-compact works, and when to use which.
Claude Code subagents: when and how to use them
Claude Code subagents explained: when to use one, three example definitions, how to pick a model and tools, and how they differ from skills and agent teams.
Claude Code usage limit reached: what to do
Hit a Claude Code usage limit? What the message means, what to do right now, why it came so fast, and how to keep working with another agent.
Run multiple Claude Code sessions at once
Four ways to run several Claude Code sessions in parallel without collisions, how to tell which one is waiting for you, and when to stop adding more.
Git worktrees for parallel AI coding agents
How to give each AI coding agent its own git worktree: the commands, what each CLI does for you, the setup traps (env files, ports, dependencies) and cleanup.
Claude Code vs Codex vs Gemini CLI (Oct 2026)
A fair comparison of Claude Code, Codex CLI and Gemini CLI as of October 2026, from their own docs: install, sign-in, permissions, undo, scripting and openness.
Which AI coding agent is best for your code?
Skip the leaderboards. Run two or three coding agents on the same real task in your repo, score the results and let your own tests decide. A simple, fair bake-off.
How to undo an AI agent’s changes safely
Snapshot before the agent starts, then undo with git restore, stash or revert. What Claude Code, Codex and Gemini CLI can rewind, and what they cannot.
Stop AI agents from editing or deleting tests
A failing test is the finish line, so an agent may edit the test instead of the code. Four layers that stop it: instructions, permissions, a hook, and review.
How to review AI-generated code fast
A ten-minute routine for reviewing AI-written code: shrink the diff, read the tests first, run the checks yourself, and read the risky paths line by line.
Is --dangerously-skip-permissions safe to use?
What --dangerously-skip-permissions, Codex --yolo and Gemini CLI YOLO mode switch off, what to try first, and how to use them without risking your machine.
Control AI coding costs and token use
Where AI coding tokens really go, how to measure usage per agent, six habits that cut it, and budgets that stop a runaway run before it drains your plan.
Run AI coding agents overnight, safely
How to let a coding agent work while you sleep without waking up to a mess: isolation, a check as the finish line, caps, no pushes, and a morning review.
Vibe coding without breaking production
A vibe-coding workflow with guardrails: work on branches, build a safety net of checks, keep secrets out of reach, ship small, and know how to roll back.
Vibe coding security: a pre-launch checklist
Ten checks before you launch an app an AI wrote: access control, secrets, settings, login, injection, payments, dependencies, logging and the agent itself.
Ship an MVP with AI agents as a solo founder
A two-week plan for a solo founder shipping an MVP with AI coding agents: a one-page spec, a thin first slice, checks, real users and what to leave for later.
Multi-agent coding workflows with approvals
Design a multi-agent coding workflow that holds up: fixed steps, one job per agent, checks between steps, and approval gates where a person has to decide.
AI coding prompts with a clear finish line
Six copy-paste prompt templates for coding agents, each ending in a check the agent can run: bug fix, feature, refactor, UI change, review and investigation.
AGENTS.md vs CLAUDE.md vs GEMINI.md
One instructions file for every coding agent: how Claude Code, Codex, Gemini CLI and Cursor read AGENTS.md, CLAUDE.md and GEMINI.md, and what to put in it.
Are MCP servers safe? 8 checks before you add one
A local MCP server is code that runs as you. Eight checks before you add one: where it runs, who made it, its version, its tool descriptions, isolation and limits.
Fix a UI bug with AI by pointing at it
How to describe a UI bug so an AI agent fixes the right thing: find the element in DevTools, capture a screenshot, write the report, then check before and after.
Voice coding: dictate to your AI agents
Four ways to talk to a coding agent instead of typing, where each one sends your audio, and a safe habit: dictate into a draft, read it, then send.
Give every agent a job you can verify
A practical way to turn vague AI coding requests into assignments with evidence and a clear finish line.
When speaking helps you code
Use voice to capture intent and feedback quickly while keeping command execution under your control.
Connect tools without losing track of access
How to think about local MCP servers, tool discovery, permissions, and review before an agent acts.