The tag view groups articles by a concrete operational problem or implementation type, without exposing hundreds of raw frontmatter labels.
Quantising Qwen3.8-27B to Q4, Q5 or W4A16 lets the model fit on a 24 GB card, but the memory you win goes on context. You choose the smallest format that passes the test on your own tasks, because our comparison changed the format, the engine and the context limit together.
A local Qwen can be reached remotely in three ways: a web interface on a private network, a private network for apps and clients, or an SSH tunnel for one engineer. Each needs an explicit secret, a description of the data flow and tests before users get access.
An AI agent chooses and carries out the next steps in a specific process, using data and tools, and hands exceptions to a person. The most common uses in companies are a shared inbox, customer service, complaints, quotes, invoices, reports, CRM, receivables, competitors, tenders, the IT helpdesk and data questions. At Syntalith an agent costs from €3,000 net. With 1,500 emails a month in a shared inbox, it pays back in 1.0 month.
A practical guide to selecting a manufacturing AI pilot by data readiness, process ownership and operational response, with office and shop-floor paths kept separate.
Compare automation and agents for repeatable work in a service company. Use your own process, review time and ongoing costs to assess a first project.
Claude Team and ChatGPT Business cost the same: a Standard seat is $25 a month or $20 with annual billing, Premium $125 or $100. So a trial on the team's tasks decides: in which tool an employee reaches a correct result with fewer corrections. At 200 analyses a month, a difference of 3 minutes is 10 hours, or €349.
You install OpenClaw with one command on macOS, Linux or Windows, and a wizard picks the model and starts the Gateway on port 18789. It needs Node.js 24.16+ or 26.1+. The MIT licence costs nothing, and at 300 tasks a month the server, the model and 2 hours of maintenance a week come to about €332 a month, of which a person's time is €302.
MCP, the Model Context Protocol, is a shared protocol through which AI assistants connect to a company's tools and data. You write an integration once on the MCP server side, and different assistants use it. Since December 2025 a foundation under the Linux Foundation has looked after the standard, and the implementation team still designs permissions and security.
n8n is a process automation platform under a fair-code licence: it links steps across apps, from fetching data to saving it and sending a notification. n8n Cloud costs from €20 a month, and the Community Edition has no licence fee, while your own server needs someone to look after it. At Syntalith an automation costs from €3,000 net, and with 150 new supplier requests a month at 20 minutes each it pays back in 2.0 months.
OpenClaw is an open-source personal AI assistant under the MIT licence. It runs on the user's device or server and links models, tools and messengers through one gateway. The code is free, and you pay for the model, the server and maintenance, about €332 a month in our scenario. In a company it makes sense when a task needs several sources checked and the next step chosen, with controlled effects.
Next step
Walk an engineer through it in 30 minutes. Within 2 business days you get, in writing, what it costs, what to do first and the starting price.
Times are shown in your own time zone. We work with clients across time zones.
See where to startOpenAI makes ChatGPT and Anthropic makes Claude. We belong to the partnership programs of both companies, so we work with their tools every day and learn from the creators themselves.
Claude Partner NetworkParticipation in programs does not mean that OpenAI or Anthropic endorses our services.