CLI tools a coding agent needs on a VPS
Install the command line tools a coding agent reaches for on a bare Ubuntu VPS, in three tiers, with the reason each one saves tokens and tool calls.
#coding-agents
Filtering by topic #coding-agents · clear
Claude effort levels: max vs high
What Claude's five effort levels really do, why high is the default, and why max burns through your Pro or Max usage window without a bigger bill.
Autoprompt skill: fewer failures, more tokens
Autoprompt solves 73 of 89 Terminal-Bench 2.1 tasks against a baseline of 60, at roughly 3x the time and 2x the tokens. What that costs on a rented VPS.
Prompt injection against coding agents
A coding agent runs on text an attacker can write. Map the real injection surfaces on a server, then rank the defences that actually shrink the damage.
Graft: a codebase map for coding agents
Graft parses your repo with tree-sitter into a codebase map your coding agent queries over MCP, so it stops rediscovering the same structure every session.
How to Write Your Own Agent Skill
Write your own agent skill by distilling one real failure: the SKILL.md anatomy, the description line that decides when it fires, and how to test it.
Why coding agents ignore your instructions
Your instruction file says stop and the agent does it anyway. The mechanics behind that, and a diagnosis you can run before you rewrite the rule again.
Multi-model routing for coding agents
Routing coding agents across models throws away the prompt cache that keeps them cheap. When routing pays, when pinning wins, and the arithmetic behind it.
Claude Max vs ChatGPT Pro: which $200 tier?
Both top tiers cost $200 a month. Compare what usage each buys, whether the coding agent is included, and how each one behaves when you hit a limit.
AI-assisted code and open source policies
Open source projects disagree on AI-assisted code. Find the project's policy before you open the pull request, and disclose it in the commit trailer.
dox: keep AGENTS.md current automatically
Your AGENTS.md is wrong three weeks later and the agent trusts it. Use dox to regenerate the file from the repo, then review the diff like code.
Coding agent telemetry: what gets sent
Four kinds of traffic leave a coding agent, and only one is unavoidable. Audit any agent from the machine itself, then cut the traffic you did not agree to.
Nested AGENTS.md files for a monorepo
One root AGENTS.md in a monorepo goes stale and burns context on directories the agent never opens. Here is the nested layout that fixes it.
Paritok token gateway: cheaper agent bills?
Paritok compresses the file reads and tool output your coding agent sends. The project claims 74% fewer tokens. Here is the mechanism and the break-even math.
How much RAM does a coding agent VPS need?
One always-on coding agent runs on 4 GB and 2 vCPU. The builds and language servers it launches are what fill the box, and what makes it hang.
MiMo-Code: Xiaomi's answer to Claude Code
What MiMo-Code actually is: Xiaomi's MIT-licensed terminal coding agent, forked from OpenCode. Read the self-reported benchmark claims before you switch.
Ponytail: the lazy senior dev agent skill
Ponytail makes a coding agent take the smallest change that works. What it ships, what its own benchmarks show, and how to copy the rule today.