525 posts tagged with "AI agents"
AI agents and autonomous systems
The Vendor Ships the Protocol, the Community Ships the Policy: What Coolify's Governed MCP Servers Teach About Agent Interfaces
Coolify's built-in MCP server covers reads while community servers add operation modes, scoped tokens, approval gates, and audit logging — a field guide to the seven governance controls every self-hosted PaaS agent interface should launch with.
E2B Ended 30-Day Sandbox Deletion — What a Self-Hosted Sandbox Should Still Do Differently
E2B quietly replaced its 30-day paused-sandbox deletion with indefinite retention — but the 1-hour and 24-hour runtime caps, the 5-minute kill-by-default timeout, and the sandbox-is-storage coupling remain. A concrete teardown of what those limits cost long-running agent workflows, plus four lifecycle policies a self-hosted sandbox should set differently.
Every PaaS Ships an MCP Server Now: Dokploy vs. Railway in the Agent-Interface Race
Dokploy mirrors all 508 API endpoints as MCP tools while Railway curates ten hosted tools plus a TypeScript deploy-your-own guide — an operation-by-operation comparison across deploy, logs, env, and rollback, and the nine-item checklist a self-hosted Render-compatible MCP surface must clear.
Only LLM Spans Should Cost Money: Span-Class-Aware Billing for Agent Observability
Span-class-aware metering prices LLM spans while tracing tool calls free — the worked cost math, which vendors already do it, and an OTel Collector recipe for self-hosted agent observability.
CNCF Says MCP Gateways Belong in the Platform. Here's What That Actually Requires
MCP gateways are now a CNCF Platform Engineering 2.0 pillar. Here's what a gateway does that an ad-hoc MCP server doesn't — OAuth 2.1 auth, per-caller rate limits, deny-by-default tool scoping, poisoning detection — and how much of it a self-hosted PaaS control plane already absorbs.
SUSE Put MCP Across Its Whole Portfolio: What a Vendor-Blessed Agent-to-Infra Interface Means for Your PaaS's Own MCP Server
SUSE embedded MCP servers across Rancher Prime and Multi-Linux Manager with AWS, n8n, Revenium, and Stacklok aboard. Here is what a vendor-blessed agent-to-infra interface validates for a PaaS agent API, where its patch-and-monitor scope stops short of the git-push deploy surface, and a checklist for building that surface yourself.
Baseten Buys Blaxel: What Sandbox-Vendor M&A Means for Your Portability Plan (and the Self-Hosted Math)
Baseten's September 2026 acquisition of Blaxel turns vendor-specific sandbox behavior into migration risk. A five-point portability checklist plus rent-vs-own math: $0.17/hr rented vs ~$0.009/sandbox-hour on a $50/mo Hetzner box.
E2B vs Modal vs Daytona in 2026: What Agent Sandboxes Get Right That Agent-Operated Deploys Still Don't
E2B and Daytona charge $0.0504/vCPU-hour for agent sandboxes while Modal charges about 40% more — but none of them hands an agent a persistent HTTPS URL with a custom domain, TLS, and rollback. A 2026 three-way comparison plus the deploy gap a PaaS still has to close.
Whose Token Is It? Solving the MCP and OAuth2 Identity Problem for AI Agent Deploys
Every deploy, rollback, or scale call an AI agent makes must answer who authorized it: the developer, the agent, or the tenant. This post maps the three-identity model, the July 2026 MCP authorization rules, and the token-exchange pattern that keeps deploy-from-chat auditable.