Bot and Agents - Talk to Wire, Skip the Terminal

Your team already has a channel they live in. The question is whether Wire can meet them there without a six-week integration project.

Wire is a command-line tool. That is fine for developers. It is a problem for the tax advisor who wants to say "clients keep asking about the new deadline" and have a page appear. The gap between Wire's capability and your team's workflow is real. The question is how big that gap actually is. Some teams close it in an afternoon. Others need it handled during onboarding. Which describes your situation?

Wire doesn't ship with a bot. That's the part that stops people. What it ships with is a Python API where every command is a clean function: `create`, `audit`, `reword`, `refine`. A Telegram bot that calls those functions is 80 lines of Python. The complexity lives inside Wire, not in the glue code. But "80 lines of Python" still assumes someone writes them. That's either you, someone on your team, or part of the managed service onboarding. Which is your situation?

Telegram, email, Slack, n8n, voice. Wire doesn't care which channel you use because it's not the channel. It's the engine underneath. Every channel calls the same commands and gets the same structured output back. The wrapper is thin by design. But here's the complication: adding a second channel later doesn't change Wire, it adds another wrapper. If your team is split across Slack and email, you're running two wrappers, not one. Does your team use one channel or several?

Every post you publish is a page you haven't written yet. A 200-character Instagram caption about shoulder pain. Seven tweets about a tax deadline. A LinkedIn article about B2B trends. Wire takes the text, researches the topic, applies your styleguide, and produces a full page. The reel gets 500 views and disappears. The page gets 500 impressions a month for years. The setup is a feed watcher: an n8n node polling your RSS, or a Zapier trigger on new posts. But there's a decision buried here. Who reviews the draft before it publishes?

The bot sorts everything into three buckets. Automatic: weekly audits, broken links, news monitoring. These run on schedule and report results without asking. Approval required: new pages, merges, full rewrites. The bot shows a preview and waits for your confirmation. Escalation: cannibalization detected, a page lost 40% of its impressions, a news development contradicts something you published. The bot flags these and explains what it recommends but won't act. Most clients start by approving everything, then loosen the rules after the first month. The boundary is yours to set. What happens when something goes wrong?

Twenty clients. Each one sends content ideas through their own Telegram bot or email address. Each bot connects to that client's Wire instance with their styleguide and their search data. The agency sees everything through one dashboard. The client never sees a YAML file or a terminal. Here's the part that matters for your decision: adding client number 11 takes an hour, not an afternoon. The bot template is reusable. Each new client is a copy with a different config file. But the first client still takes an afternoon to set up. Is that the managed service onboarding or something you're building yourself?

Wire is `pip install wire`. The bot layer is a setup step, not a product you buy. Managed service clients get the bot configured during onboarding: Telegram by default, email or Slack if your team prefers. The approval rules, the schedule, the channel. All part of the €1,200 onboarding. Self-service users set up their own wrapper. Agencies replicate the pattern per client. The first client takes an afternoon. Each one after takes an hour. The question isn't whether this is possible. It's whether you're setting it up or someone is setting it up for you.

Wire is a command-line tool. That is an implementation detail, not a requirement. The people who benefit most from Wire — the tax advisor, the mechanic, the agency account manager — should never see a terminal.

They talk to a bot. The bot talks to Wire. Wire does the work.

How It Works

A Telegram bot is the simplest version. You send a message:

"Clients keep asking about the new Grundsteuer deadlines. We should have a page on that."

The bot understands this is a content request. Behind the scenes, it runs wire.content create guides/grundsteuer-deadlines with your styleguide already loaded. Wire searches the web, writes the page in your voice, lints it, and deploys it. The bot sends you a preview link.

"The brake repair page sounds too technical. Customers say 'Bremsen schleifen' not 'Bremsbelagverschleiß'."

The bot passes this as a styleguide refinement. Wire updates the terminology, rewrites affected pages, and shows you the diff.

"What happened this week?"

The bot triggers wire.chief audit and returns the health report: 3 pages need news updates, 1 cannibalization pair detected, 12 new keyword opportunities. You tap "fix all" or review each one.

Not Built In — Set Up Once

Wire does not ship with a Telegram bot or a Slack integration. What it ships with is a Python API that any wrapper can call. The bot layer is a separate piece — part of the onboarding when you use the managed service, or something you wire up yourself in an afternoon.

The reason this is easy and not a consulting project: Wire's commands are already clean single-purpose functions. create, audit, reword, refine — each one takes a topic and a slug and returns a result. A Telegram bot that calls these functions is 80 lines of Python. An n8n webhook that triggers an audit is a single HTTP node. The complexity lives inside Wire, not in the glue code.

During onboarding, we set up the channel that fits your team. After that, the bot runs unattended. You talk to it. It talks to Wire. No maintenance, no updates to the wrapper — when Wire gets better, the bot gets better automatically.

Telegram Is Just the First Channel

The pattern works the same everywhere. Wire is the brain. The channel is the mouth and ears.

Telegram bot. Direct messages or a group chat where your team sees what Wire does. Quick replies, inline previews, approval buttons. Best for solo operators and small teams who live in Telegram already. Setup: one bot token from BotFather, one Python script, 30 minutes.

Email. Forward a customer question to wire@yourdomain.com. Wire turns it into a draft page and sends it back for review. Best for professionals who work from their inbox and don't want another app. Setup: one forwarding rule, one script that reads the inbox.

n8n / Make / Zapier. Wire exposes its commands as webhook endpoints. An automation platform connects triggers (new support ticket, CRM tag, calendar event) to Wire actions (create page, run audit, generate report). Best for agencies managing multiple client sites with existing automation stacks. Setup: one webhook per command, drag-and-drop in n8n.

Slack. A /wire slash command in your workspace. Team members request content without leaving their chat. Best for companies where content is a team responsibility, not one person's job.

Voice. A phone call or voice memo transcribed and routed to Wire. "Heute hat ein Kunde gefragt, ob man Winterreifen auch im Sommer fahren darf. Mach eine Seite dazu." Wire gets the intent, does the research, writes the page. Best for tradespeople who don't type. Setup: a transcription service (Whisper, Google Speech) feeding into the same bot script.

Every channel uses the same Wire underneath. Switching from Telegram to Slack or adding email on top does not change how Wire works. You add a new wrapper. Wire doesn't notice.

Every Instagram post, tweet, or LinkedIn update you publish is a seed. You already created the thought — it just lives on a platform that won't rank on Google and disappears from feeds within hours.

Wire can watch these channels and turn posts into pages.

A physiotherapist posts an Instagram reel explaining why shoulder pain gets worse when sleeping on your side. The caption is 200 characters. Wire takes that caption, researches the topic, and produces a full page with your styleguide applied — your terminology, your perspective, your clinical experience. The reel gets 500 views and disappears. The page gets 500 impressions per month for years.

A tax advisor tweets a thread about the Elster deadline changes for 2026. Seven tweets, each 280 characters. Wire combines them into a structured guide with proper headings, internal links to related pages, and the keywords people actually search for. The thread gets 40 retweets. The page captures search demand that exists 12 months a year.

An agency posts a LinkedIn article about B2B lead generation trends. Wire detects it via RSS or webhook, evaluates whether it fits the client's site, and drafts a page. The agency reviews and publishes with one tap.

The flow: social post → Wire intake → research → full page → review → publish. Your social media stops being ephemeral and starts being a content pipeline input. Every post you've already written is a page you haven't published yet.

The setup is a feed watcher — an n8n node that polls your Instagram RSS, or a Zapier trigger on new tweets. When a new post arrives, it forwards the text to Wire's create command. That is the entire integration.

What the Bot Handles

Not everything needs a human decision. The bot sorts incoming requests into three categories:

Automatic. Weekly audits, broken link fixes, news monitoring, internal linking. Wire runs these on schedule. The bot reports results but doesn't ask permission. You set the schedule once during setup.

Approval required. New pages, merges, full SEO rewrites. The bot shows a preview and waits for your "yes." Nothing publishes without your confirmation.

Escalation. Cannibalization detected between two important pages. A page lost 40% of its impressions. A news development contradicts something you published. The bot flags these and explains what it recommends, but won't act until you decide.

You configure the boundary between automatic and approval-required once. Most managed service clients start conservative — approve everything — and loosen it after the first month once they see Wire's output quality.

The Agent Layer

Under every channel sits the same agent: a thin translation layer between natural language and Wire commands. The agent does three things:

Interpret. Parse what the user wants. "We need something about X" becomes create. "This page is wrong" becomes improve. "What's happening" becomes audit.
Execute. Call Wire with the right parameters. The agent knows the site's topic structure, the styleguide, the current audit state. It picks the right command and the right scope.
Report. Translate Wire's output back to the user's language. Not "MERGE: 2 pairs (weaker page absorbed into stronger)" but "Two of your pages were competing for the same Google searches. Wire combined them. Here's what changed."

The agent is stateless. It doesn't remember conversations. Wire's database remembers everything: every audit, every keyword snapshot, every page version. The agent just connects human intent to Wire's capabilities.

Building this agent is not a large project. Wire's commands return structured data. The agent wraps them with a Claude call that translates intent in and results out. One prompt, one function, done.

For Agencies: One Bot, Many Clients

An agency running 20 client sites doesn't give each client CLI access. Each client gets their own Telegram bot (or email address, or Slack channel). Each bot connects to that client's Wire instance with their styleguide and their GSC data.

The agency sees all clients through one dashboard. Wire runs the pipeline per site. The bots handle client communication. The agency's value is curation and strategy — deciding which suggestions to approve, which topics to prioritize, when to create new pages.

Client input flows naturally: "Kunden fragen neuerdings viel nach Wärmepumpen-Förderung" arrives via Telegram. The agency reviews the draft Wire produces. One tap to publish. The client never sees a YAML file, a terminal, or a git commit.

Setup per client: one Wire instance, one bot, one styleguide session. The bot template is reusable — each new client is a copy with a different config file. Adding client number 11 takes an hour, not an afternoon.

What You Need

Wire is pip install wire. The bot layer is not a product you buy — it is a setup step.

Managed service clients get the bot configured during onboarding. Telegram is the default. If your team prefers email or Slack, we set that up instead. The bot, the schedule, the approval rules — all part of the €1,200 onboarding.

Self-service users set up their own wrapper. A Telegram bot is a free bot token + 80 lines of Python calling Wire's API. An n8n integration is a webhook + a few nodes. The getting started guide covers Wire itself. The bot wrapper is documented separately because it depends on which channel you pick.

Agencies replicate the pattern per client. The first client takes an afternoon. Each subsequent client takes an hour. Wire handles the content operations. The bot handles the communication. Your time goes into strategy, not plumbing.

The Point

Your expertise runs through channels you already use. Wire never requires you to open a terminal, write markdown, or understand what "keyword cannibalization" means. You describe what your clients need in your own words. Wire turns it into content that ranks.

The CLI exists for builders and power users. The bot exists for everyone else. Both run the same engine. The bot is not a separate product — it is a thin layer that takes 30 minutes to set up and never needs touching again.