Orchestrator
The Orchestrator is your command center for talking to specialized Valdr agents. Ask a question in plain language, select the agent that owns the job, and keep its structured answer and follow-up work in one conversation thread.
Use it to inspect current state and take supported actions without translating every request into individual tool calls yourself.
What this page covers
- Orchestrator layout: threads, chat area, and controls
- Orchestrator agents: Gunnar, Mimir, Nikol, Skadi, and Verdandi
- Launcher presets for provider configuration
- Thread management and conversation history
- Real-world examples of orchestrator workflows
Quick start
1 — Open the Orchestrator
Click Orchestrator in the sidebar navigation. The chat interface loads with the threads panel on the left and the message area on the right.
2 — Pick your agent
Use the Orchestrator agent dropdown at the bottom to select the right agent for your question. Each agent specializes in a different domain.
3 — Ask your question
Type a natural-language question in the input box and click Send. The agent processes your request using real project data and returns a structured, formatted response.
4 — Continue the conversation
Follow up in the same thread to drill deeper, or click + New to start a fresh conversation.
Orchestrator agents
The orchestrator routes your questions to specialized agents — each one tuned for a specific domain of project operations. Select the agent that matches what you need.
| Agent | Domain | What to ask |
|---|---|---|
| Gunnar Orchestrator | Project management & task creation | Sprint status, task breakdowns, project health, creating and assigning tasks |
| Mimir Knowledge Orchestrator | Workspace knowledge | Knowledge sources, indexed docs and code, symbol relationships, and agent memory |
| Nikol Registry Orchestrator | Agent, prompt, and capability management | Agent configurations, prompt composition, capability assignments, registry operations |
| Skadi Sprint Orchestrator | Sprint lifecycle management | Sprint planning, shaping, staffing, status checks, closing sprints |
| Verdandi Workflow Orchestrator | Durable workflows | Workflow authoring, run inspection, diagnosis, and recovery |
Gunnar — General project management
Gunnar is the default orchestrator and handles the broadest range of questions. Ask about project health, task status, sprint history, or have Gunnar create new tasks and plans. Gunnar sees your full project context — tasks, sprints, agents, and their relationships.
Example prompts:
- What is the status of our sprints?
- Create a task to add dark mode support
- What tasks are in progress right now?
- Summarize the work completed this week
Mimir — Workspace knowledge
Mimir curates and searches workspace knowledge. Use Mimir when an answer depends on indexed documentation, code, symbol relationships, or a learning that should be available to future agents.
Nikol — Agent and registry operations
Nikol specializes in managing your agent registry, prompts, and capabilities. Use Nikol when you need to configure agent behavior, review prompt composition, or manage capability assignments without leaving the chat.
Example prompts:
- List all bot agents in the registry
- What capabilities does the auditor agent have?
- Show me the prompts linked to the executor agent
- Create a new agent for frontend code review
Skadi — Sprint management
Skadi owns the sprint lifecycle. From planning and shaping sprints to staffing and status checks, Skadi gives you structured visibility into your delivery cadence.
Example prompts:
- What is the status of our sprints?
- Create a new sprint for the authentication feature
- Which sprints closed in the last 7 days?
- Staff the current sprint with reviewers
Verdandi — Workflow operations
Verdandi authors, runs, inspects, debugs, and repairs durable workflows. Use Verdandi when you need to create a workflow definition, understand why a run stopped, or choose a targeted recovery action. Learn more about Verdandi.
Layout and controls
The orchestrator interface has three main areas:
Threads panel
The left panel manages your conversation history:
- + New — Start a fresh conversation thread
- Search — Filter threads by keyword to find past conversations
- Thread list — Each thread shows the agent name, status (open/closed), provider, timestamp, and session ID
- Collapse — Hide the threads panel to maximize the chat area
Each thread is a self-contained conversation with full history. Threads are labeled with the orchestrator agent name and show whether the session is open or closed.
Chat area
The main panel displays the conversation between you and the orchestrator agent. Agent responses render with full Markdown formatting — headings, tables, code blocks, and lists are all supported. Each response includes a copy button for quick extraction.
Bottom controls
Two dropdowns control how messages are processed:
- Launcher preset — Selects the AI provider and model configuration (e.g., Claude GPT OSS). Presets are configured in Settings.
- Orchestrator agent — Selects which specialized agent handles your messages. Switch agents mid-session to ask different types of questions.
The text input and Send button complete the interface. Type your question, press Enter or click Send, and the selected agent processes it.
Thread management
Creating threads
Click + New to start a fresh conversation. The new thread inherits your currently selected launcher preset and orchestrator agent. Each thread runs as an independent agent session.
Navigating threads
Click any thread in the panel to load its conversation history. Threads display:
- Agent name and partial session ID
- Open or closed status
- Provider used (e.g., “claude”)
- Relative timestamp
Searching threads
Use the Find thread… search box to filter by keyword. This searches across thread metadata to help you locate past conversations quickly.
Deleting threads
Click the trash icon on any thread to remove it. This deletes the thread and its conversation history.
Launcher presets
The launcher preset dropdown determines which AI provider and configuration backs the orchestrator agent. Presets define:
- The AI provider (e.g., Claude, OpenAI-compatible endpoints)
- Model selection and parameters
- API key configuration
Presets are managed in Settings and can be shared across your team via Provider Packs.
Tip
Configure multiple launcher presets to test different models or providers with the same orchestrator agent. Switch between them without changing anything else.
Why this matters
- One command center for orchestrated work: Ask questions, create tasks, check status, and coordinate supported agent operations from a shared view.
- Structured, actionable answers: Orchestrator agents return formatted responses with tables, lists, and next steps — not walls of text. Copy what you need and keep moving.
- Grounded project context: Agents can query your actual Valdr sprints, tasks, agents, workflow runs, and history instead of relying only on the current prompt.
- Specialists from official packs: After importing the relevant packs and configuring a launcher preset, choose Gunnar, Mimir, Nikol, Skadi, or Verdandi for the job each one owns.
- Conversation history you can search: Every thread is saved with full context. Come back to a conversation days later and pick up where you left off.
Next steps
- Multi-Agent Workflow Orchestration — Understand the platform capability behind orchestrator agents
- Agent Sessions — Inspect the full execution trail behind any orchestrator conversation
- Agents — Browse and configure the orchestrator agents in the registry
- Sprints — Manage sprints directly from the Sprints view
- Tasks — Work with tasks created through the orchestrator