Skip to content
nexus-agents

Try nexus-agents with the tour

In this tutorial you install nexus-agents and step through its built-in tour. By the end you will have seen what four of its tools return: task routing, a consensus vote, research synthesis and audit-chain verification.

The tour prints stored example output. It makes no model calls, so you need no API keys and no authenticated coding CLI. Everything it shows is labelled representative output; your own runs will differ.

You need Node.js 24 and npm.

  1. 1 Install nexus-agents

    npm install -g nexus-agents
nexus-agents --version

    You should see the installed version on the last line:

    nexus-agents v11.1.3

    Your version number may be newer. If the shell says command not found, npm’s global bin directory is not on your PATH; see Installation.

  2. 2 Start the tour

    nexus-agents tour

    You should see the tour banner and the first step:

    ═══ nexus-agents tour ═══
…
─── Step 1/5: Welcome ───

    After each step the tour waits with:

    Press Enter to continue (Ctrl-C to exit) ...

    Press Enter to move on.

  3. 3 Read the routing step

    Step 2 shows the orchestrate tool. Look at the block between --- representative output --- and --- end output ---:

    Task: "implement a /healthz endpoint with structured logging"
  Classifier:    task_type=implementation  complexity=4/10
  Router picked: claude-sonnet           (capability=code_generation, score=0.91)
  Pattern:       graph                    (2 sequential subtasks detected)
…

    Notice the three decisions it records: what kind of task this is, which model was picked, and which execution pattern was chosen.

  4. 4 Read the vote step

    Step 3 shows a three-voter consensus vote:

    Proposal: "Adopt Bun as the dev-time test runner alongside Vitest"
  Software Architect   APPROVE  (conf 0.82)  Keeps deps slim; vitest API parity
  Security Engineer    REJECT   (conf 0.74)  New supply-chain surface, unproven
  Scope Steward        REJECT   (conf 0.71)  Two runners = sprawl; pick one
  Result: 1/3 approve (33%)  ->  REJECTED  (simple_majority, threshold 50%)

    Each voter gives a decision, a confidence and a one-line reason. The result line names the bar the tally was measured against. One approval out of three is below the 50% bar, so the proposal is rejected.

  5. 5 Finish the tour

    Keep pressing Enter through step 4 (research synthesis) and step 5 (audit-chain verification). The tour ends with:

    ═══ Tour complete ═══
Next steps:
  nexus-agents doctor              -- check your install
  nexus-agents setup               -- configure MCP + .rules + data dirs
  nexus-agents --help --all        -- see every command

    To print the whole tour again without the pauses, run:

    nexus-agents tour --non-interactive

What you did

You installed nexus-agents and read example output from routing, voting, research and audit verification, without spending any model quota.

Next