> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.moveworks.com/harness/agentic-orchestrator/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.moveworks.com/_mcp/server.

# Agentic Orchestrator

## Docs

- [Artifacts](https://docs.moveworks.com/harness/agentic-orchestrator/artifacts.md): Artifacts are durable files, such as reports, spreadsheets, decks, images, and PDFs, that the Agentic Orchestrator returns in the conversation.
- [Planning](https://docs.moveworks.com/harness/agentic-orchestrator/planning.md): How the Agentic Orchestrator decides whether to answer, call tools directly, or run a plan the user reviews first.
- [Code Execution](https://docs.moveworks.com/harness/agentic-orchestrator/code-execution.md): The Agentic Orchestrator runs Python in a sandbox to compute exact results over data and files, or to generate a file for the user.
- [File Processor Plugin (Agentic Orchestrator)](https://docs.moveworks.com/harness/agentic-orchestrator/file-processor.md): On the Agentic Orchestrator, the assistant answers questions about an uploaded file by reading it with the File Processor plugin. No button, no mode, nothing to exit.
- [Intelligence Core](https://docs.moveworks.com/harness/agentic-orchestrator/intelligence-core.md): The Agentic Orchestrator's model roles and what the reasoning model sees on each call.
- [Memory](https://docs.moveworks.com/harness/agentic-orchestrator/memory.md): What the Agentic Orchestrator keeps in context during a conversation, how sessions resume, and how long state is kept.
- [Multilingual](https://docs.moveworks.com/harness/agentic-orchestrator/multilingual.md): The Agentic Orchestrator works in your employees' languages natively, with no translation layer in between.
- [People Graph](https://docs.moveworks.com/harness/agentic-orchestrator/people-graph.md): How the Agentic Orchestrator answers questions about colleagues, managers, and reporting lines with People Lookup.
- [Skills](https://docs.moveworks.com/harness/agentic-orchestrator/skills.md): Package instructions and supporting files that the Agentic Orchestrator loads on demand when a request matches.
- [Convert an Existing Skill to the Agentic Orchestrator](https://docs.moveworks.com/harness/agentic-orchestrator/skills/convert-a-skill.md): Adapt a skill bundle written for another agent host so it runs on the Agentic Orchestrator.
- [Tool Orchestration](https://docs.moveworks.com/harness/agentic-orchestrator/tool-orchestration.md): Which tool frameworks the Agentic Orchestrator can call, how tools reach the model, and what happens to each result.


> **Note:** This page contains both a page directory (above) and the landing page content (below). The page directory is generated for agent use and does not appear on the landing page.

> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.moveworks.com/harness/agentic-orchestrator/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.moveworks.com/_mcp/server.

# Agentic Orchestrator

> **Controlled Availability**
>
> The Agentic Orchestrator is in Controlled Availability and comes to your Enterprise Assistant in Q1 2027. Capabilities, limits, and interface labels described here may change before general availability. To request access, contact your Moveworks account team.

Compared with the [Agentic Reasoning Engine](/harness/agentic-reasoning-engine), the Orchestrator pares back the constraints on the model and gives it new tools: Skills, code execution, artifacts, and plans the user approves. The loop itself is unchanged, a model reasoning over the request and calling tools until the work is done, so an assistant can take on longer and more open-ended work.

```mermaid
flowchart TD
    U["User message"] --> M["Model<br />instructions, Skills, memory"]
    M -->|"multi-step work"| P["Plan<br />user approves"]
    M -->|"call tools"| T["Tools and code"]
    P --> T
    T -->|"results fold back"| M
    M --> R["Reply and artifacts"]
```

## What the Upgrade Adds

* **Instructions on every turn.** Your instructions are applied verbatim on every turn, after the Orchestrator's fixed base prompt.
* **Skills.** Uploaded bundles of procedures and reference material that load only when a request matches their description. A Skill tells the model how to work; a Tool Studio plugin gives it a tool to call.
* **One loop for every tool.** MCP servers, A2A agents, and Tool Studio plugins run in the same reasoning loop as built-in search.
* **New built-in capabilities.** Code execution, artifacts such as reports, spreadsheets, and decks, and file and image understanding. [Code Execution](/harness/agentic-orchestrator/code-execution) covers uploaded files, and [Intelligence Core](/harness/agentic-orchestrator/intelligence-core) covers images.
* **Plans for longer work.** Multi-step work goes through a plan that the user reviews and approves before it runs, and the session suspends while it waits on the user.

## Where It Runs

Today the Orchestrator runs Productivity Assistants, which you build in Assistant Builder in the EmployeeWorks web app (**Browse assistants**); see [Create a Productivity Assistant](/ai-assistant/productivity-assistants/create-a-productivity-assistant). Your Enterprise Assistant moves to it in Q1 2027.

## Topics

#### [Artifacts](/harness/agentic-orchestrator/artifacts)

Reports, spreadsheets, decks, and images returned as files.

#### [Planning](/harness/agentic-orchestrator/planning)

Answer directly, call tools, or draft a plan the user approves.

#### [Code Execution](/harness/agentic-orchestrator/code-execution)

Sandboxed Python over uploads and Skill resources.

#### [File Processor](/harness/agentic-orchestrator/file-processor)

Answers questions about uploaded files. No button, no mode.

#### [Intelligence Core](/harness/agentic-orchestrator/intelligence-core)

The models inside the harness and what each one does.

#### [Memory](/harness/agentic-orchestrator/memory)

What the model sees on each turn and how sessions persist.

#### [Multilingual](/harness/agentic-orchestrator/multilingual)

How the Orchestrator works across languages.

#### [People Graph](/harness/agentic-orchestrator/people-graph)

People Lookup over your organization's people data.

#### [Skills](/harness/agentic-orchestrator/skills)

Packaged procedures that load on demand, and how to configure them.

#### [Tool Orchestration](/harness/agentic-orchestrator/tool-orchestration)

Plugins, MCP servers, A2A agents, and built-in tools, and how results are processed.