How much does Odin AI cost?

Pro is $20 per seat a month with credit usage pooled across the team. Platform actions such as document uploads, chat messages and workflow runs cost flat credits, and model calls cost token credits, at one cent per credit. Optional one time Accelerator packages add an Odin build team: Growth at $15,000 for two custom AI employees deployed in 30 days, and Scale at $30,000 for four in 60 days. On premise licenses carry no public price.

Does Odin AI support MCP and tool calling?

Full under the Agentic Index verification standard. Odin AI publishes public evidence of Model Context Protocol (MCP) support, tool calling, and API or SDK extensibility for connecting agents to external tools. Evidence basis: A REST API spans projects, knowledge bases, agents, models, roles, users, teams, tools, chat, chat tests, documents, meetings and analytics, so a developer can create and edit custom agents, add files or URLs to a knowledge base and run cha. Graded from public evidence only against the Agentic Index 14 point capability taxonomy. No vendor pays for placement.

Does Odin AI have SOC 2, SSO and role based access control?

Full under the Agentic Index verification standard. Odin AI publishes public evidence of enterprise security and identity controls, covering a compliance posture such as SOC 2 or ISO 27001, single sign on (SSO or SAML), and role based access control (RBAC). Evidence basis: SOC 2 Type II and ISO 27001 certifications and HIPAA compliance back the platform, data is encrypted in transit with TLS 1.2 or later and at rest with AES 256, each customer's documents and workflow data stay isolated, and customer data doe. Graded from public evidence only against the Agentic Index 14 point capability taxonomy. No vendor pays for placement.

Can Odin AI be self hosted or run in a private VPC?

Full under the Agentic Index verification standard. Odin AI publishes public evidence of self hosted or private VPC deployment, along with data residency options for controlling where agent data is processed and stored. Evidence basis: Two delivery models are on offer. The cloud service runs in four AWS regions, US East in Northern Virginia, London, Mumbai and Sydney, each with a static egress IP that customers add to their allowlists. An on premise license puts the whole. Graded from public evidence only against the Agentic Index 14 point capability taxonomy. No vendor pays for placement.

Back to vendors

Odin AI

Also known as: Odin, Odin AI Platform, getodin.ai

Visit site

Agent builderindependentVerified 2026-10-18

Entry pricePro $20 per seat a month, with credit usage pooled across the teamFull pricing detail

Odin AI is a no code enterprise platform for building AI agents and automations that act across more than 200 business apps, grounded in a company's own knowledge base. Agents get toolkits, memory, MCP servers and manager and worker delegation, workflows start on schedules, emails, file uploads or chat, and the platform runs in Odin's cloud or on premise.

Odin AI, operated by One Realm, Inc., sells an agent and automation platform to enterprises in banking and insurance, healthcare, telecom, manufacturing and business process outsourcing, with a support agent, a chatbot builder and an AI notetaker built on the same base. The company claims more than 200 enterprise deployments, and Automation Anywhere runs its support email through Odin, which resolves more than 2,700 tickets a month across 15 locations.

An agent combines a personality, a chosen model, persistent memory and toolkits for knowledge search, web search, documents, SQL databases and Smart Tables, Python and Node.js sandboxes, WhatsApp, Telegram and Notion, plus connectors to Microsoft 365, Gmail, Google Drive, Salesforce, HubSpot, Jira, Confluence and Asana and any MCP server. A manager agent delegates to worker agents in router, pipeline, fan out, hierarchical or committee patterns.

The Automator and Workflow Builder chain agents, tools, conditions and outputs on a visual canvas, started by chat input, a file upload, a schedule or an incoming email, with version history and per step run logs, and Autopilot turns one prompt into a tested, deployed workflow. Tools can require a person's approval before they run, the Ask User toolkit pauses an agent for structured answers, and memory learns user preferences and facts that wait for approval before an agent uses them.

Pro costs $20 per seat a month and includes unlimited agents, Autopilot, workflows, the integrations, API access and webhooks, bring your own key models, automatic model routing and fallback, SAML single sign on and audit logs, with usage metered in credits worth one cent each. Fixed price Accelerator packages add Odin engineers who build AI employees to order: Growth at $15,000 for two in 30 days and Scale at $30,000 for four in 60 days.

Odin holds SOC 2 Type II and ISO 27001 certifications and HIPAA compliance, encrypts data with TLS 1.2 or later and AES 256, and does not train models on customer data. The cloud service runs in four AWS regions (Northern Virginia, London, Mumbai and Sydney), and an on premise license runs the whole platform in a customer's own cloud account or data center on Kubernetes.

Odin fits enterprises in regulated industries that want business teams to build governed agents over their own systems and data, with the option to keep everything in house. It fits less well for teams that need agents to operate a browser or desktop, or an evaluation suite that scores agent changes against test sets before release.

Vendor details

Canonical URL

https://getodin.ai

Subcategory

No code enterprise agent and workflow automation platform with knowledge grounding, multi agent delegation and an on premise option

Funding status

Privately held; the platform is operated by One Realm, Inc.

Company status

independent

Use cases & customers

Primary use cases

customer and internal support agents grounded in a knowledge basedocument intake such as invoice extraction, validation and sync to accountingemployee onboarding across HR, IT and facilities with approvalsprospect research and Salesforce lead enrichmententerprise search and question answering over company filesmulti agent workflows that update CRM, ERP, support and collaboration tools

Target customers

banks, insurers and financial services teams automating KYC, lending, claims and fraud workhealthcare administration teams handling intake, referrals, claims and authorizationstelecom service, provisioning and billing operationsBPO and shared services centers running email triage, finance, HR and QA workmanufacturers automating supplier, quality, warranty and plant operationsIT and security teams that need AI running inside their own infrastructure

Deployment options

SaaS in four AWS regions: US East (Northern Virginia), London, Mumbai and SydneyOn premise or private cloud license on customer Kubernetes, cloud account or data center

Integrations

More than 200 integrations through the Automator, with agent connectors for Microsoft 365 mail, calendar, OneDrive and SharePoint, Gmail, Google Drive, Sheets and Slides, Excel, Word, Dropbox, Asana, Jira, Confluence, Salesforce and HubSpot, toolkits for WhatsApp, Telegram, Notion and SQL databases, and any MCP server with OAuth or custom headers. Chatbots publish to websites, Slack, Microsoft Teams and Google Chat, the support agent connects to Zendesk and Intercom, and a REST API and TypeScript SDK cover chat, Smart Tables and voice. Models come from OpenAI, Anthropic, Google, Meta Llama and DeepSeek, or through your own keys for Azure OpenAI, AWS Bedrock, OpenRouter and custom endpoints.

In practice

An accounts payable team builds an Automator workflow that starts when an invoice email arrives, extracts the fields, checks them against policy rules, sends exceptions to a reviewer for approval and syncs approved invoices to the accounting system, with each run's step logs kept for audit.

A telecom support group loads its help articles and policies into a knowledge base, publishes a support agent in Microsoft Teams and on its website, and routes sensitive account questions to a human team while the agent answers routine ones with cited sources.

A bank runs Odin on premise in its own Kubernetes cluster, signs staff in through Okta over SAML, and has a manager agent hand KYC document checks to a worker agent that runs Python validation in a sandbox, with the tool set to wait for a compliance officer's approval before the result is written back.

Sources & related URLs

Related / legacy domains

https://learn.getodin.ai https://docs.getodin.ai https://getodin.ai/pricing/ https://getodin.ai/faq/

Research sources

Agentic Index coverage score

13.0 / 14 capabilities · 93%

Integrations & Tool Calling	Full
Agents reach business systems through connectors for Microsoft 365 mail, calendar, OneDrive and SharePoint, Gmail, Google Drive, Sheets and Slides, Excel, Word, Dropbox, Asana, Jira, Confluence, Salesforce and HubSpot, and the Automator adds more than 200 integrations and toolkits for CRMs, support desks, databases, HR and finance tools. Any MCP server attaches with OAuth 2.0 or custom headers, its tools prefixed by server name, with quick add tiles for popular servers such as Sentry. Toolkits run SQL queries, manage Smart Tables, send WhatsApp and Telegram messages and execute Python or Node.js in isolated sandboxes. An agent can create a Jira issue, update a Salesforce lead or send a Microsoft 365 email as part of its run. Sourcelearn.getodin.ai/agents/external-integrations/overviewread 2026-10-18
Workflow Orchestration	Full
With the Agent Communication toolkit, one agent becomes a manager that delegates to specialized worker agents, passes conversation context along, merges their outputs and falls back to another agent or a person when a worker fails, in router, pipeline, parallel fan out, hierarchical or expert panel patterns. Workflow Builder V2 lays out agents, tools, conditions and final outputs on a drag and drop canvas, and Automator flows mix AI extraction, deterministic rules, approvals and app actions in a single run. Autopilot takes one prompt, selects the toolkits, wires the integrations, tests the run and deploys the workflow once it passes. Sourcelearn.getodin.ai/agents/toolkits/agent-communicationread 2026-10-18
Knowledge Grounding & RAG	Full
Each project's knowledge base takes uploaded files and folders in PDF, Word, Excel, PowerPoint, text and CSV formats, single URLs, website crawls and documents pulled from Google Drive and SharePoint, and agents search it through the Knowledge Base toolkit. Admins choose sentence or semantic chunking and the chunk settings, attach custom metadata and resource types to documents, and resync a file to reprocess it. Answers carry citations to their source documents, the history view shows how many documents shaped each response, and Smart Tables hold structured data that agents query alongside the text. Sourcelearn.getodin.ai/knowledge-base/kb-overviewread 2026-10-18
Human Oversight & Guardrails	Full
Any tool can be set to require approval, so the agent asks a person to confirm before it sends, updates or deletes anything, and the Ask User toolkit pauses a run to put structured questions on screen as choices or free text before it continues. Automator flows place human review steps and approval paths in front of sensitive actions and route exceptions to people, and the support agent hands sensitive tickets to a human team. Guardrails hold answers to ingested company data and flag responses below a confidence threshold, and admins can whitelist the domains an agent's web tools may visit. Sourcelearn.getodin.ai/agents/toolkits/ask-userread 2026-10-18
Security, Identity & Governance	Full
SOC 2 Type II and ISO 27001 certifications and HIPAA compliance back the platform, data is encrypted in transit with TLS 1.2 or later and at rest with AES 256, each customer's documents and workflow data stay isolated, and customer data does not train models. Projects start with Admin, Editor and Viewer roles, and custom roles set view self, view all or edit access per feature, separately from team level roles. SAML 2.0 single sign on, included in Pro and configured from the Super Admin dashboard on premise, works with Okta, Entra ID, Ping, ADFS, OneLogin, Auth0, Keycloak and Google Workspace and is keyed by email domain, so removing someone at the identity provider cuts their access. Sourcelearn.getodin.ai/general/security-complianceread 2026-10-18
Observability & Auditability	Full
Project settings hold an audit log of every action, with the user, time, action type, the user's role at that moment and the credits consumed, searchable, filterable and exportable for compliance reporting. The Agent Interactions tab records each message with its agent, model, tokens, tool call count and duration, and expands to the full tool calls and trace details. Analytics track messages, users, response times, feedback on correct and incorrect answers and LLM metrics, workflow runs keep per step logs, and owners get an email when a workflow fails. Sourcelearn.getodin.ai/project-settings/logsread 2026-10-18
Memory & State Persistence	Full
When memory is on, an agent reads each user message in the background for preferences, facts such as role or team, and useful background, checks each new item against what it already holds, gives it a confidence score from 0 to 1 and saves it as pending. Pending memories expire after seven days unless someone approves them, and approved ones enter the system prompt for that user's later conversations, sorted into preferences, user information and important context. Approval is configurable for teams that want automatic saving, and memories stay scoped to the individual user rather than shared with everyone who talks to the agent. Sourcelearn.getodin.ai/agents/memoryread 2026-10-18
Deployment & Data Residency	Full
Two delivery models are on offer. The cloud service runs in four AWS regions, US East in Northern Virginia, London, Mumbai and Sydney, each with a static egress IP that customers add to their allowlists. An on premise license puts the whole platform, agents, automations, knowledge base and audit controls, inside a customer's own cloud account, VPC or data center on enterprise Kubernetes, which keeps prompts, documents, logs, embeddings and workflow outputs in that environment, while the customer's team runs upgrades, sizing and day two operations with Odin's guidance and support. Sourcelearn.getodin.ai/offerings/on-premise/overviewread 2026-10-18
Prebuilt Agents / Templates / Packs	Full
Ten ready agent templates arrive configured with behaviors and toolkits: knowledge base, data analyst, research assistant, project manager, Python developer, business intelligence, prospect research, vendor onboarding, Salesforce lead enrichment, and an agent builder that drafts new agents. Knowledge base templates, a prompt library and portable Agent Skills folders package procedures for reuse across agents, and Autopilot assembles a whole workflow from a single prompt. Fixed price Accelerator packages, where Odin's engineers build custom AI employees to order, sit apart from this self serve catalog. Sourcelearn.getodin.ai/agents/agent-templatesread 2026-10-18
Triggers & Channel Coverage	Full
Workflows start from chat input, a file upload, a schedule or an incoming email, and Automator flows also take webhooks and pick up files, messages, forms, tickets and records from connected systems. Agents and public chatbots answer on a website widget and inside Slack, Microsoft Teams and Google Chat, and the support agent covers Zendesk, Intercom, WhatsApp and Telegram. Outside applications start chats and voice conversations through the REST API and the TypeScript SDK, so the same agent can serve employees in their chat tool, customers on the website and back office systems through code. Sourcelearn.getodin.ai/tools/workflows-v2read 2026-10-18
Model Flexibility & Routing	Full
Predefined models from OpenAI, Anthropic and Google, along with Meta's Llama 3 and DeepSeek, work without keys, and Pro adds automatic routing and fallback between models. Bring your own keys connects Azure OpenAI, OpenAI or Anthropic compatible endpoints, the Google AI API, OpenRouter, AWS Bedrock and private custom endpoints, and each agent runs on whichever model it is assigned. Odin's own Hoshi 1.0 model, built for tool use and structured output, costs fewer credits than frontier models and answers through an OpenAI compatible Chat Completions API. Sourcelearn.getodin.ai/agents/modelsread 2026-10-18
APIs / SDKs / MCP Extensibility	Full
A REST API spans projects, knowledge bases, agents, models, roles, users, teams, tools, chat, chat tests, documents, meetings and analytics, so a developer can create and edit custom agents, add files or URLs to a knowledge base and run chats from outside the app. The TypeScript SDK packages ChatSDK for streaming chat interfaces, SmartTablesSDK for structured data and VoiceSDK for voice conversations with React hooks, authenticated with server side API keys or access tokens for web apps. Workflows export and import between projects, and agents extend through MCP servers and Agent Skills. Sourcelearn.getodin.ai/sdk/overviewread 2026-10-18
Testing, Debugging & Optimization	Partial
Chat tests pair a question with an expected answer, sit in named test groups and run as message groups through the API, which gives a team a repeatable set of prompts to replay after a change. Workflow Builder V2 adds per step run logs for debugging and a version history for restoring earlier versions, Autopilot tests a generated workflow before deploying it, and thumbs up, thumbs down and written feedback feed a chart of correct and incorrect answers. Test runs come back as transcripts, with no automatic scoring against the expected answers or comparison across agent versions. Sourcedocs.getodin.ai/reference/create_chat_test_chat_create_test_postread 2026-10-18
Browser / Computer-use	Partial
The Get Content tool visits a URL during a run and renders it in a headless browser, so an agent can read sites built with JavaScript, with ads and navigation stripped away and clean text returned. Web Search queries Google or Bing for titles, snippets and links, and administrators can restrict both tools to whitelisted domains. Neither tool clicks, types, logs in or fills in forms, and no agent operates a desktop. Work in other systems runs through API connectors, toolkits and MCP tools, so a portal that needs a login or a form submission stays out of reach. Sourcelearn.getodin.ai/agents/toolkits/web-searchread 2026-10-18

The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded

Pricing

Pro $20 per seat a month, with credit usage pooled across the team

Per seat monthly subscription, plus credits for platform actions and model tokens at one cent per credit

Included quota

Pro covers unlimited AI agents and workflows with usage pooled across the team; the number of credits included per seat is not set out.

What is public

The Pro seat price and its features, both Accelerator package prices with scope and timelines, and the credit rules and value of a credit.

Billing mechanics

Self serve per seat subscription billed to the team owner, with credit consumption tracked per project, member and action in a billing log.

Cost watchouts

Large document uploads and long contexts consume token credits on top of seat fees, and premium models cost more credits per call than Odin's Hoshi model. Done for you builds cost $15,000 or $30,000 one time through the Accelerator packages.

Variable cost rationale

Seats are a fixed $20, but every upload, message, tool call and model response draws credits, and token credits rise with document size, response length and premium model choice, so heavy knowledge base use or frontier models raise the monthly bill.

Additional watchouts

Credit use scales with document volume and model choice, so teams loading large knowledge bases or running frontier models spend more than the seat price suggests.

Overage / add-ons

Usage beyond the pooled allowance is billed in credits: flat credits per platform action, such as 1 credit per 10,000 words uploaded and 1 credit per workflow run, plus token credits at the model's input and output rates, each credit worth $0.01.

Sales call required

No, self serve available

Free / trial

No free plan with set limits; Pro is the entry plan

Lowest paid plan

Pro at $20 per seat a month: unlimited agents, Autopilot, workflows, 200+ integrations, BYOK models, SAML SSO and audit logs

Commercial notes

Accelerator packages add a dedicated implementation engineer and account manager, workflow design and build, integration setup and post launch monitoring on top of Pro access. On premise runs on a separate license issued by Odin.

Key ambiguities

How many credits each Pro seat includes before pooled usage is billed.

Missing data

Credits included per Pro seat and the price of the on premise license.

Agentic Index verified 2026-10-18

Official pricing page Docs

Data confidence: high