Building an AI agent on Talona takes about five minutes. You describe what you want the agent to do in plain English, the platform assembles the right skills and integrations, and you deploy it to a chat channel of your choice. This guide walks through the full path from empty prompt box to a Telegram bot answering messages from your team.
What you will need
Before you start, have these ready:
- A Talona account. The free tier is enough to follow this guide. See plan details on the pricing page.
- A Telegram account, if you want to deploy the agent to Telegram. Other channels (Slack, web chat, email) follow the same pattern.
- A clear sentence describing what your agent should do. The more specific, the better the result.
Step 1: Describe the agent
Open the Talona landing page and type a sentence describing the agent you want to build. The builder uses your description to pick skills, integrations, and tools, so concrete descriptions produce better agents than vague ones.
Examples of good descriptions
- "An agent that reads my GitHub notifications every morning and posts a digest to my team Slack."
- "A research assistant that takes a topic, browses the web, and returns a one-page brief with citations."
- "A support agent that answers questions about our pricing using the docs in this Notion workspace."
Examples of weak descriptions
- "A general-purpose AI assistant."
- "Something to help with productivity."
The first set will produce an agent with the right tools wired up on the first try. The second set will require several rounds of clarification and still produce something generic.
Step 2: Review the generated config
Within a few seconds of submitting your description, the builder streams back a draft configuration showing the model, skills, integrations, and tools it has selected. This is the moment to make corrections. The most common edits are:
- Add or remove an integration (e.g. swap Notion for Confluence).
- Tighten the system prompt with specifics about your team or use case.
- Add a skill from the catalog the builder did not pick up automatically.
You can iterate by typing a follow-up message. The builder updates the config in place and shows a diff so nothing is silently overwritten.
Step 3: Connect a channel
Open Settings and pick a channel. Telegram is the fastest to wire up if you do not already have a workspace integration to test with.
Telegram setup
- Open Telegram and start a chat with @BotFather.
- Send /newbot and follow the prompts to name the bot. BotFather will return a bot token.
- Paste the token into Talona under Settings → Channels → Telegram.
- The bot starts answering messages immediately. No redeploy required.
Step 4: Test the agent
Send the agent a message that exercises one of its tools. If you built a research assistant, ask it to summarize a recent news topic. If you built a support agent, ask it a question that should land in the docs you connected. Watch the trace viewer in the dashboard to see which tools fired and how long each step took.
If the agent picks the wrong tool or returns a vague answer, the fix usually goes in the system prompt or a skill rather than a code change. Open the builder, describe what you want different, and the config updates without losing your channel wiring.
Common questions during the first deploy
Most first-deploy stumbles happen at the channel step. If your agent does not respond to the first message on Telegram, check that the bot token was pasted in full and that you have started a chat with the bot from your account, not the workspace owner's account.
If the agent responds but ignores the integration you added, open the trace viewer and check whether the integration was actually called. The most common cause is an integration that needs an OAuth approval that has not been completed yet. The integrations page shows a red dot next to anything pending.