Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Choose tools for your use case

Start with a problem, then add tools when you need them. These are suggested workflows, not additional Tambo profiles or a promise that every connection is configured for you. Tambo installs and verifies a selected set; indexing, agent registration, event sources and policies still need their own setup.

A developer working with a coding assistant

Goal: find the right code and carry useful context between tasks.

Start with: Tambo’s retrieval profile: Quipu for durable knowledge, Camayoc for checked knowledge ingress and Bobbin for code search/context. tambo quickstart is the local, single-user route. Index the repository you want to work on and register its server as described in the agent guide.

Example workflow: search for the code behind a behavior, make a change, then preserve its rationale and source as knowledge for the next task.

Add when needed: Yupana through code-intel for callers and change impact; Desire Path to capture failed tool calls rather than repeating them. Policy checking needs configured policies; an index alone is not enforcement.

Next: getting started, using agents, profiles.

An engineering team or agent-crew lead

Goal: keep ownership, progress and handoffs visible across several workers.

Start with: Seeds for work items and Shantytown for a host-based agent crew. The crew profile includes the knowledge/retrieval tools and the Chaski, Seeds and Shuttle members. Choose the runtime explicitly, for example tambo plan --profile crew --crew shantytown, then review the plan before apply.

Example workflow: assign a tracked task, let an agent work on it, record the result with provenance and use a signed workflow for a handoff that needs an explicit transition. Configure Chaski reactions for the events you want delivered.

Add when needed: Rikuy for a shared live brief; Yupana for code-policy and impact checks. A live brief needs its data sources, and notifications need configured subscriptions. Neither follows automatically from installing tools. The optional creel, both and standalone runtime choices are described in profiles; both retains explicit handoff rather than automatic fan-out.

Next: profiles, crew contracts, the stack.

A knowledge curator or researcher

Goal: retain evidence and vocabulary so a claim can be traced and revisited.

Start with: the kg profile, Quipu plus Camayoc. Quipu keeps provenance and history; Camayoc prepares and checks ingress. Quechua supplies the shared vocabulary and shapes that help tools agree on the meaning of facts.

Example workflow: prepare a source-backed episode, validate its terms and provenance, admit it through the configured gate, then query the recorded facts and their history. Choose the write rules you need rather than assuming every installation refuses the same inputs.

Add when needed: reviewed corpus shares, or Chaski to notify you when relevant knowledge changes. Shares are staged imports; Tambo does not automatically promote them into ROOT. Repository-pendant availability is a separate publication gate.

Next: the corpus, profiles, Quipu’s book.

A platform or service operator

Goal: connect service changes, delivery records and operational decisions.

Start with: Quipu and Camayoc for source-backed operational knowledge and delivery ingress. Tambo supplies a reviewable plan and verification; start small with kg and review additional members as your workflow grows.

Example workflow: configure a supported delivery emitter, record a deployment outcome, query its change history and use that evidence in an operational decision.

Add when needed: Chaski for event-driven notifications, Shuttle for signed runbooks and Rikuy for status and decisions in a live brief. Configure each event source, subscription and view explicitly. Tambo does not enable deployment emitters or prove DORA coverage; missing history is not a measured zero.

The everything profile is a broader tool selection, not an all-services production deployment. Check the generated plan and each tool’s proof boundary.

Next: the stack and its proofs, profiles, Camayoc’s delivery catalog.

An individual organizing projects and decisions

Goal: keep tasks and the reasons behind decisions together.

Start with: Quipu and Camayoc for durable notes with provenance; add Seeds when the work needs tracked items. A crew plan with --crew standalone makes the tools available without installing a crew runtime.

Example workflow: record a decision and its source, create the related work item, then update its state as the work progresses. Query the evidence again when you revisit the decision.

Add when needed: Chaski for configured reminders or notifications, and Rikuy for a live brief. This is a knowledge-and-work workflow, not a built-in calendar or personal-planning profile. No agent crew is required.

Next: profiles, using Tambo, the stack.

Turn a recommendation into an installation

The available profile names are kg, retrieval, code-intel, crew and everything. The canonical stack table describes the twelve projects; it is not an install manifest. In particular, everything does not mean all twelve projects are installed. Inspect the plan’s actual members, review compatibility, and retain the plan and state paths so you can resume or verify the same installation.