Give your AI agent its own identity
An AI agent using a copy of your API key is invisible in the audit log. Give it its own OIDC client in authentik, scope its token, and revoke only that one.
#ai-agents
Filtering by topic #ai-agents · clear
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.
Host your AI-generated app on a VPS
Your agent wrote the app. Now keep it running: systemd supervision, a reverse proxy with TLS, an env file for secrets, backups, and safe redeploys.
What is Claude Code?
Claude Code is Anthropic's terminal coding agent. What it does, where it runs, what a plan costs, and why it is not a chat window or an IDE.
Claude Code output styles explained
An output style edits Claude Code's system prompt, so it shapes every reply. What the built-in styles do to your transcript, and how to write your own.
Self-host OpenBot AI coworkers on a VPS
Self-host OpenBot so each AI coworker gets its own container and browser. How the gateway decides every action, and what that costs you in RAM.
Self-host OneCLI: one agent per person
OneCLI hands every person a sandboxed agent and keeps the keys in one gateway. Self-host it on a VPS: Docker Compose, PostgreSQL, and real sizing numbers.
What is an agent harness?
An agent harness is the program around the model: the loop, the tools, the permissions and the session state. How it differs from a model and a framework.
Self-host Deer Workflow: agent graphs on a VPS
Deer Workflow puts agent orchestration in reviewable TypeScript. Install it on a VPS with Bun, pin the version, and run one graph headless under systemd.
Self-host Moli, a headless browser for agents
Headless Chrome is the biggest process on a small VPS. Install Moli, serve CDP on loopback, point an agent at it, and learn where it stops working.
DeepSeek Harness vs Claude Code vs Omnigent
DeepSeek Harness, Claude Code and Omnigent compared on architecture, model coupling, licence and maturity, plus what each one takes to run on a VPS.
Agent memory poisoning explained
Prompt injection ends with the session. Memory poisoning does not. How hostile text becomes a stored fact, and how to isolate, expire and review it.
Reach the DSH web UI on your VPS
The DSH web UI binds to 127.0.0.1:3080, so the URL your VPS prints will not open on your laptop. Three safe ways to reach it, and one you should avoid.
DeepSeek Harness plugins worth installing
Five DeepSeek Harness plugins that change how a rented server behaves: budget caps, tool permission rules, injection scanning, durable memory, LAN access.
Run an SEO audit agent on your VPS
The seo CLI turns a spare VPS into an SEO audit box: a systemd timer runs the crawl, and your agent queries the results over a local MCP server.
Old Coder skill: review evidence, not code
Your agent writes a SPEC you approve and an EVIDENCE report you can rerun yourself. Inside the Old Coder gauntlet: what mutation testing adds over coverage.
Tura: an agent harness that cuts tokens
Tura claims 77.5% fewer tokens than Codex CLI. Here is how the harness gets there, and how to check that claim against your own provider bill.
Self-host the SandBase agent runtime
Run SandBase Harness v0.3.2 on your own VPS: tagged install, agent YAML, MCP servers, sandbox modes, and pointing the Anthropic SDK at your box.
Gate AI agent actions behind approvals
An agent proposes an action. A policy service and a person decide it. A sealed executor holds the only credentials. The design that survives prompt injection.
Self-host Rakazo: your own AI bot on a VPS
Run Rakazo on your own VPS: Node 22, pnpm, Postgres and Graphile Worker in Docker Compose, the sandbox provider choice, key handling, and honest sizing.
Claude Code Hooks Explained
Claude Code hooks run whether the model agrees or not. Where they live, which events fire, what exit code 2 does to a tool call, and the security cost.
Run a headless browser for AI agents on a VPS
Headless Chromium on a VPS breaks in known ways: tiny /dev/shm, sandbox flags, missing fonts, leaked processes. Set the limits before your agent hits them.
Self-host open-kritt AI security scanning
Run open-kritt on a VPS: Docker Compose setup, pinning a release, SSH tunnel to the UI on port 5173, and the provider budget to set before your first scan.
Self-host LiveContext, an n8n alternative
LiveContext CE is a six container Docker stack with a Java backend that wants 8 GB. Size it, pin the version, put Traefik in front, and back up both stores.
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.
Can your VPS run Firecracker microVMs?
Firecracker needs /dev/kvm, and most VPS plans never pass it through. Check yours in three commands, read the result, and know what to run when it is missing.
Claude Code auto mode: the new default
Auto mode becomes the Claude Code default on 14 August 2026. What each permission mode does, and which one a server you cannot watch deserves.
Claude Code sessions that message each other
One Claude Code session can send text to another on the same VPS. What ListAgents and SendMessage do, when a second session helps, and why messages get held.
Stateless MCP servers: what actually changed
MCP revision 2026-07-28 removed sessions and the initialize handshake. What that means for your reverse proxy, health checks, timeouts and auth.
Self-host KiroCrew on a VPS: always-on agent
Run KiroCrew as a pinned container on your own VPS so memory and schedules survive reboots. Docker, systemd, SSH access, backups and rollback.
The types of AI agents, explained
Simple reflex, goal-based, utility-based, learning, multi-agent: what each type of AI agent is for, and which ones you can honestly self-host.
Self-host OpenTag for @agent mentions
Run OpenTag on a VPS so Slack and GitHub mentions reach your coding agent: TLS ingress, webhook signature checks, token scopes, and safe defaults.
Self-hosted evals for AI agents
Build an eval loop you own: golden cases from real traces, cheap deterministic checks first, an LLM judge second, and a pass rate you track per commit.
The Fable method: skills for any model
The fable-method repo turns Claude Fable 5's habits into agent skills. What each file does, what ports to other models, and how to A/B it on a VPS.
Reach your Hermes agent from your phone
Your Hermes agent runs on a VPS. Reach it from a phone over a private mesh or a public HTTPS endpoint, and revoke access when the phone goes missing.
Self-hosted stock research agent on a VPS
Build a self-hosted stock research agent on a VPS: a market data feed, a DuckDB store, a systemd timer that fires at market close, and an LLM screen.
Self-hosted Firecrawl alternatives for a VPS
Draco, Hound and self-hosted Firecrawl compared on RAM, headless browser needs and API compatibility, with a pinned install and an MCP wire-up.
MCP email server: give your agent an inbox
Run an MCP email server on your VPS so Claude can triage your inbox. App password scoping, sender allowlists, draft-only replies, and the injection risk.
Self-host sandboxd: an AI app builder on a VPS
Run the sandboxd AI app builder on your own VPS: pinned install, model keys, HTTPS preview URLs, RAM and disk floors, and cleaning up stale sandboxes.
Self-host Dormice for agent sandboxes
Dormice runs E2B-compatible agent sandboxes on one VPS you own. Install it, execute code in a sandbox, check the isolation, and size the host.
AI Agent vs LLM vs AI Assistant
An LLM is weights that need RAM. An assistant adds a chat surface. An agent adds tools and a loop, and it holds credentials, so it needs a box that stays on.
What are agent skills, really?
An agent skill is a folder with a SKILL.md file that loads only when your request matches it. Why that beats one giant prompt, and how it differs from MCP.
Self-host an AI PR review agent on a VPS
Run an AI pull request reviewer on a VPS you control: self-hosted runner, diff-scoped prompts, path and size filters, inline comments, and the cost per PR.
Omnigent: one harness for many agent CLIs
Omnigent drives the agent CLIs you already have installed. What a meta-harness is, how to pin release 0.7.0, and how to sandbox each sub-agent on a VPS.
Give your AI agent SearXNG web search
Your SearXNG instance can be the search backend for an AI agent. JSON API setup, the trust boundaries, and the prompt injection surface it opens.
DESIGN.md: the file after AGENTS.md
AGENTS.md tells a coding agent how to work in your repo. DESIGN.md tells it why the code is shaped that way, so it stops undoing your decisions.
Numbat: watch what your AI agents do
Perplexity open sourced Numbat, an endpoint monitor that records what AI coding agents really do on your servers. What it sees, and what it cannot stop.
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.
Self-host Open Connector for AI agents
Run the Open Connector auth gateway on your own VPS so your agents never hold a SaaS token: pinned image, TLS origin, OAuth callbacks, backups.
How to learn AI agents, from scratch
A staged path for learning AI agents: concepts, a loop you write yourself, tools, memory, and safety, with one thing to build at every stage.
Memmy: a local memory hub for agents on a VPS
Memmy gives your AI agents one shared memory store. Build it from source on Ubuntu, run the memory service on port 18960, and keep every note local.
tty7: a terminal workbench for agents
tty7 is a Rust terminal workbench with a session daemon, native SSH and agent-aware panes. Install it, then see where tmux on the VPS still wins.
n8n AI agent: build one on your own VPS
Build a working AI agent in n8n: the AI Agent node, a Claude model credential, an HTTP Request tool, memory, a trigger, and the settings that cap cost.
What is loop engineering? A definition
Loop engineering means designing the trigger, boundary, verification and budget an AI agent repeats, instead of writing one clever prompt. A plain definition.
Recall: give Claude Code a memory on a VPS
Recall is a fully local plugin that logs each Claude Code session and condenses it into a resumable summary. Set it up on a VPS and measure the saving.
Keep secrets out of your AI agents
An agent holding your API keys can leak them in one tool call. Give it scoped short lived tokens behind a credential gateway, never the real keys.
AGENTS.md and HUMAN.md, explained
AGENTS.md is the README your coding agent reads. What belongs in one, what never does, how CLAUDE.md fits, and a starter template to copy.
Give coding agents a disposable VM
AI coding agents belong on a machine you can destroy. Blast radius, clean state per task, snapshots, and the VPS pattern that keeps it cheap.
Claude for Finance Analysis: a Practical Guide
Use Claude to analyze stocks and options with live market data: a free no-key API, three ways to connect it, and copy-paste prompts that get grounded answers.
Best self-hosted AI agents in 2026
OpenClaw, Hermes, Dify, OpenHands, and Agent Zero compared: what each agent does best, what it needs, and how to run any of them safely on your own VPS.
Run OpenCode on a VPS
OpenCode is the most starred open source coding agent. Install it on a VPS, run it in tmux as an unprivileged user, and keep its API key locked down.
Run Claude Code safely on a server
Claude Code can run any command your user can. What the skip permissions flag changes, and how to contain the blast radius, from sandbox to disposable VPS.
Self-host Dify on a VPS with Docker Compose
Six containers, so budget 4 GB of RAM. Replace every secret in .env before you boot it, then create the admin account at /install before a stranger does.
Build your own OpenClaw-style AI agent
OpenClaw is a personal AI agent that runs commands on your own server. Here is how one is built, piece by piece, and why hardening comes first.
Hermes Agent hosting: a $5 VPS is enough
Hermes needs no GPU: the model runs elsewhere, so a small VPS is plenty. Run it as its own user with ProtectSystem=strict in systemd, and cover IPv6 in UFW.
Run MCP servers on a VPS for AI coding agents
Run Model Context Protocol servers on your own VPS so AI agents get real tools: stdio and remote HTTP transports, systemd, TLS, auth, and every failure mode.
Build your own AI agent on a VPS
An AI agent is a loop around a language model that can use tools. Learn the concepts behind building one on your own VPS: the loop, tools, MCP, memory.
Build an AI agent with Claude on a VPS
Use Claude as the brain and your VPS as the body. How the Messages API, tool use, and MCP fit together to build an AI agent you run yourself.
Run a coding AI agent on a VPS
Run Aider or Goose on a VPS to code from your terminal, connected to a hosted or a self-hosted Ollama model. Setup, model sizing, and safety.
Run OpenClaw safely on a VPS
OpenClaw runs shell commands and browses the web, so a careless setup is dangerous. Harden it on a VPS: unprivileged user, firewall, secrets, systemd.
Self-host OpenHands on a VPS
OpenHands is an AI agent that writes and runs code, so its setup needs care. Install it on a VPS with Docker and lock down the Web UI and its secrets.
Self-host Agent Zero on a VPS
Agent Zero runs code, a browser, and a shell from a Web UI, so exposing it carelessly is dangerous. Set it up on a VPS and lock the Web UI down.