Moveworks

What it does

Writes code

PartlyThe Code Interpreter plugin writes Python to compute answers over live plugin data; Agent Studio produces agent configuration, not application source code.The Code Interpreter Plugin can write code to analyze data related to a user's request and provide accurate & verifiable calculation.docs.moveworks.com/ai-assistant/capabil… ↗

Debugs and fixes

PartlyIt diagnoses misconfigurations in the agents you build in Agent Studio, not failures in your own codebase.Agent Architect helps you review your builds, resolve misconfigurations and performance gaps, and iterate until every agent is reliable, compliant,…moveworks.com/us/en/platform/ai-agent-b… ↗

Reviews code

PartlyIt reviews agent builds it helped assemble, not a diff written by another engineer.Validate your agents end-to-end and catch issues before they reach production.moveworks.com/us/en/platform/ai-agent-b… ↗

Ships it

PartlyIt deploys the AI agents built in Agent Studio, including promotion between instances; it ships no application code.A purpose-built, low-code workspace to create, configure, and deploy proactive and responsive AI agents at the speed of business, complete with…moveworks.com/us/en/platform/ai-agent-b… ↗

Done-for-you service

PartlyYou license software, but Moveworks Forward Deployed Engineers will implement, integrate and build custom solutions under a Statement of Work.Moveworks Professional Services (PS) is your partner for strategic initiatives, new implementations, and custom solutions that go beyond the scope of…docs.moveworks.com/ai-assistant/getting… ↗

How much it does unattended

Runs autonomously

YesAmbient agents run on webhook or schedule triggers with no user prompt, interpreting the event and executing actions; builders insert approvals where they want a human.Ambient agents monitor enterprise systems continuously, detecting events or changes to initiate workflows automatically.docs.moveworks.com/agent-studio/core-co… ↗

Multi-agent

YesThe reasoning engine selects and executes several plugins in parallel on one request and merges their outputs; ambient agents can also hand off to conversational agents.The Moveworks Assistant calls multiple plugins in parallel, reviews the responses from them, and combine their individual responses into a single,…docs.moveworks.com/agent-studio/agentic… ↗

Agent permissions

YesAgents act with the user's own OAuth-delegated access, and builders gate actions with the Approvals Engine (sequential, parallel, conditional) and Policy Validators.Enable AI agents to act securely on behalf of users through OAuth 2.0, giving employees full visibility and control to manage access.moveworks.com/us/en/platform/ai-agent-b… ↗

Which models it runs on

Claude

NoThe Model Reference enumerates only OpenAI gpt-* and o-series models; Anthropic is reachable only as a customer-built HTTP Action calling the Anthropic API with your own key.Specifies the LLM model to use. See the Model Reference for available options. Defaults to gpt-5.4-minidocs.moveworks.com/agent-studio/actions… ↗

GPT

YesGPT foundation models do the reasoning and action planning, and LLM Actions default to gpt-5.4-mini with the whole gpt/o-series selectable.Foundation Models, e.g. GPT 5.4, 5.5, etcdocs.moveworks.com/agent-studio/agentic… ↗

Open models

NoNo open-weight model is selectable: every model in the Model Reference is a proprietary OpenAI model served through OpenAI Direct or Azure OpenAI.Specifies the LLM model to use. See the Model Reference for available options. Defaults to gpt-5.4-minidocs.moveworks.com/agent-studio/actions… ↗

Model choice

PartlyA builder picks the model per LLM Action from a fixed OpenAI list; the Assistant's own reasoning models are chosen and routed by the Moveworks LLM Gateway.we have built an LLM Gateway, a single egress point that dynamically routes requests from other internal Moveworks services to the best machine…docs.moveworks.com/agent-studio/agentic… ↗

Where you use it

In your editor

NoDeployment surfaces are chat platforms and the browser; Agent Studio is a hosted low-code IDE, not an extension for a code editor.1. Slack 2. Microsoft Teams 3. Google Chat 4. Webex 5. Zoom Chat 6. Embedded AI Assistantdocs.moveworks.com/ai-assistant/chat-pl… ↗

On the command line

NoNo terminal surface is published: the supported surfaces are chat platforms and the web, and every listed developer tool is browser-based.1. Slack 2. Microsoft Teams 3. Google Chat 4. Webex 5. Zoom Chat 6. Embedded AI Assistantdocs.moveworks.com/ai-assistant/chat-pl… ↗

In your pipeline

PartlyAn ambient agent can be triggered by GitHub webhooks including pull_request events and act through HTTP Actions, but there is no CI integration or PR review surface of its own.GitHub webhooks notify your endpoint of repository events like pushes or pull requests.docs.moveworks.com/agent-studio/system-… ↗

In a browser

YesAssistant on Web runs the full assistant in any browser, and the Embedded AI Assistant can be dropped into ServiceNow, SharePoint, Google Sites or Unily.It’s the same trusted Assistant employees already use in chat messaging platforms like Slack and Microsoft Teams, now accessible directly from any…docs.moveworks.com/ai-assistant/ai-assi… ↗

Whose machine it runs on

Self-hosted

NoThe platform runs in Moveworks' AWS cloud; the only customer-hosted component is a container agent that proxies calls to on-premises systems.Communication between the Agent and the Moveworks platform is always initiated out-bound only by the Agent to the Moveworks Cloud Platform.docs.moveworks.com/agent-studio/core-pl… ↗

Open source

NoThe only Moveworks-published source is the Content Gateway starter server on GitHub, an integration scaffold; the platform itself ships as a hosted service.The Content Gateway Starter Code gives you a working Python server that implements the full Content Gateway protocol.docs.moveworks.com/ai-assistant/enterpr… ↗

What it costs to run

How it meters

PartlyThe one published figure meters per user per year on a 12-month contract; moveworks.com names no unit at all and routes buyers to a custom quote.1,000-2,500 users | Price per user per year | $150.00aws.amazon.com/marketplace/pp/prodview-… ↗

Free tier

NoNo free tier, trial or developer instance: you build inside a tenant your account team provisions, and moveworks.com only offers a demo or a quote.You will need to build in your organization's tenant. Please request access for a sandbox instance from your Moveworks admin or account team.docs.moveworks.com/ai-assistant/getting… ↗

API access

YesBearer-authenticated REST APIs (Events, Conversations, Data) plus webhook listeners, with a separate base URL per data centre.Moveworks provides REST APIs to integrate AI-powered automation into your applications and workflows.docs.moveworks.com/api-reference/overvi… ↗

MCP server

PartlyMoveworks is an MCP client: admins connect external MCP servers in MCP Workspace, which is in Controlled Availability for a limited set of customers. No MCP server exposes the assistant to other agents.The MCP client is the runtime that lets the Moveworks assistant discover and call those servers' tools in a conversation, using the user's…docs.moveworks.com/agent-studio/mcp-wor… ↗

Bring your own key

PartlyYou cannot bring your own inference for the reasoning engine, but a plugin can call an external LLM API with your own credentials held in a connector.Set up an HTTP action to call the external agent's API. Here's an example using the Anthropic API:docs.moveworks.com/agent-studio/cookboo… ↗

Buying it for a team

A company can buy it

YesSold only to organisations, either as a quoted enterprise package on moveworks.com or as a 12-month contract through the vendor's AWS Marketplace listing.Scalable packages and pricing options for all stages of business growthmoveworks.com/us/en/pricing ↗

Seat model

PartlyThe published dimension is per user per year with a 1,000-user floor and a 2,500-user ceiling; above that band pricing reverts to a quote.1,000-2,500 users | Price per user per year | $150.00aws.amazon.com/marketplace/pp/prodview-… ↗

Admin controls

YesRole-based access scopes which tools and data an agent may touch, connectors are closed by default, and Policy Validators check rules before an action executes.Provide users with defined access to specific tools and data, ensuring control, and preventing unauthorized or accidental changes.moveworks.com/us/en/platform/ai-agent-b… ↗

Audit log

YesA JSON SIEM export over SFTP, refreshed every three hours, covers authentication, permission changes, config changes and Agent Studio actions.Moveworks provides a JSON-based SIEM log export (via SFTP) that captures API calls, authentication events, permission changes, config changes, and…docs.moveworks.com/agent-studio/core-pl… ↗

Single sign-on

YesSAML or OIDC per SSO config, with connectors for Okta, Azure AD, Ping and OneLogin and multiple IdPs supported on one tenant.Choose protocol (SAML or OIDC)docs.moveworks.com/ai-assistant/ai-assi… ↗

Not established: Pooled budget

What happens to your code

Opt out of training

YesCustomer data is excluded from global generative model training by default, so there is nothing to opt out of.No customer data is used to train global generative models.moveworks.com/us/en/platform/security ↗

Data residency

YesHosted in AWS US Commercial, EU, US GovCloud, Canada, Australia, UK and Japan, each with its own API base URL, and data stays in region.All customer data remains in its region and is never transferred elsewhere.moveworks.com/us/en/platform/security ↗

Getting out

PartlyPublished plugins export to another tenant (connectors stripped, never across data centres) and analytics pull out through the Data API; no conversation-data export or deletion terms are published.When exporting plugins from one tenant to another, we drop the connector by design from the plugin.docs.moveworks.com/agent-studio/develop… ↗

Certifications

YesISO/IEC 42001, 27001, 27017, 27018 and 27701, SOC 2 Type 2, CSA STAR Level 2, GDPR, CCPA and FedRAMP are all claimed on the trust page.Strict compliance with global, regional, and industry-related privacy laws and security standards, including FedRAMP® and GDPR.moveworks.com/us/en/platform/ai-agent-b… ↗