Your first hour with Hive

This walkthrough is the fastest way to understand how Hive’s parts fit together. Complete it with disposable example work before connecting production systems.

0–10 minutes: finish setup

1. Launch Hive and complete the environment checks.
2. Open Models.
3. For local use, connect Ollama and pull a model that fits your available memory.
4. For cloud use, configure one provider key and verify its live model list.
5. Open Settings and confirm the Hive home, authentication, Docker, Git, and GitHub status you intend to use.

Expected result: at least one model is selectable in an agent editor.

10–20 minutes: meet Staff and create an agent

1. Open Staff and inspect a recipe-seeded profile.
2. Review its role, prompt, skills, tools, memory, suggestions, performance, and history tabs.
3. Open Agents → New Agent → Staff profile.
4. Select that profile and create its agent.
5. Open the agent’s editor and review the inherited runtime settings.

Expected result: the Staff profile has an assigned agent, and the agent card shows a model.

20–30 minutes: chat and refine

1. Select Chat on the agent card.
2. Give it a small, verifiable request.
3. Start a second session for a different topic.
4. Ask it to remember one harmless preference.
5. Edit the agent, open Memory, and confirm the saved text.
6. Update one prompt instruction and retry the original request in a fresh session.

Expected result: you can distinguish conversation history, durable memory, and the system prompt.

30–40 minutes: build a pipeline

1. Create a second agent with a different responsibility.
2. Open Pipelines → New Pipeline.
3. Add a first step using {input}.
4. Add a second step using {prev}.
5. Review Input Flow Preview.
6. Save and run the pipeline manually.

Expected result: the second agent receives the first agent’s output and produces a final result.

40–50 minutes: found a Colony team

1. Open Colony and choose a small recipe.
2. Found a team with a durable name.
3. Review the proposed crew and their Staff profiles.
4. Add a narrowly scoped work item.
5. Launch and watch blackboard updates and handoffs.

Expected result: you can see the difference between a predetermined pipeline and a role-driven Colony mission.

50–60 minutes: automate safely

1. Open Schedules.
2. Create a disabled schedule targeting the tested agent or pipeline.
3. Write the same prompt that already worked manually.
4. Run it once manually and inspect history.
5. Decide whether to enable it.

Do not expose a webhook or enable a consequential schedule during the tour.

Where to go next