--- title: "Orangescrum AI Chat - Talk to Your Projects" description: "Chat with your Orangescrum workspace in plain English. Ask about blockers, log time, run sprint reports - powered by MCP and your real project data." canonical: https://www.orangescrum.com/ai-chat --- # Orangescrum AI Chat - Talk to Your Projects > For the complete documentation index, see [llms.txt](https://www.orangescrum.com/llms.txt). Powered by MCP - Connected to your real data # Chat with your projects. Get real answers. Ask about tasks, blockers, team workload, and sprint progress in plain English. Orangescrum AI doesn't guess - it reads your actual project data. [Start for free](/sign-up)[See it in action](/ai-chat#prompts) Orangescrum AI Connected to your workspace Today, Monday 9:42 AM You What's blocking Sprint 12? 9:42 AM·Seen Reading Sprint 12 from your workspace... AI Sprint 12 has **3 open blockers** as of today: #OS-441API rate limit issueBlocked RMRavi M.Due 3 days ago #OS-437Auth token expiry bugIn Review PKPriya K.Due tomorrow #OS-429Mobile layout regressionBlocked UnassignedDue today Want me to reassign #OS-429 or create a priority flag for Ravi? 9:42 AM·Read from live data AI AI is typing... ⌘K Powered by your live workspace dataPress Enter to send Who it's for ## Built for how your team actually works Three different roles, one AI that speaks all of them fluently. ### Team Members Stop hunting through tickets. Ask what you're working on, [log time](/time-tracking) instantly, and stay unblocked - all from a single chat. - See your assigned tasks at a glance - Log hours without opening a timer - Get context on any task, instantly ### Project Managers Get a real-time pulse on every [sprint](/agile-project-management). Ask what's overdue, who's overloaded, and what's at risk - without digging through [reports](/project-reports-analytics). - Sprint health summaries on demand - Identify bottlenecks before they escalate - Workload view across the whole team ### Developers Build on top of it. Orangescrum AI is [MCP-native](/mcp) - bring your own model, embed it in your tools, or extend it with custom integrations. - Any LLM provider supported - Full MCP protocol access - Embed in any workflow or UI What you can ask ## Plain English. Real answers. No commands, no filters, no exports. Just ask. Task status ### “What did I complete this week?” You closed 7 tasks this week across 2 projects. Highest priority: #OS-412 is done and in review. Blockers ### “Are there any blockers in Sprint 12?” 3 tasks are blocked. #OS-441 is overdue by 3 days and unassigned - want me to flag it? Time logging ### “Log 3 hours on the API integration task” Done. Logged 3h on #OS-437. Your total this week is now 18h 30m. Overdue work ### “Which tasks are past their due date?” 5 tasks are overdue. 2 are high priority and unassigned. I can create a summary report for your standup. Sprint summary ### “Give me a sprint 11 wrap-up” [Sprint 11](/agile-project-management): 34 tasks completed, 4 carried over. Velocity: 82 points. Team avg: 6.2 tasks/member. Team workload ### “Who has capacity this week?” Ravi and Priya have bandwidth. Meera is over-allocated by 12h. Want me to suggest a [rebalance](/resource-management)? How it works ## Up in minutes. Useful forever. 1 ### Connect via MCP Orangescrum AI plugs into your workspace through the [Model Context Protocol](/mcp) - reading your real projects, tasks, sprints, and timelogs live. 2 ### Ask in plain English No dashboards, no filters, no exports. Type what you want to know or do. The AI understands context, not just keywords. 3 ### AI acts on real data Get precise answers - or take actions like [logging time](/time-tracking), [updating task status](/task-management), and flagging blockers - all without leaving the chat. Permissions ## You control who can chat with your data. Role-based access keeps AI usage aligned with how your organization works. Admins set it once - it applies across your entire workspace. **Owner & Admin** always have full access - no toggle needed. **Users, Clients, and Guests** can be enabled or restricted independently per role. **Read vs. write** permissions are separate - a Client can ask questions without ever modifying a task. **Tenant-wide kill switch** lets admins disable AI for the entire company instantly. RoleReadWriteAI Chat OwnerAlways AdminAlways User Client Guest Write permissions require confirmation for **every action**. The AI never modifies data silently. Rate Limits ## Guardrails that keep costs predictable. Set company-wide and per-user caps on messages and tokens. No runaway bills, no surprise overages. Daily company cap 2,000 messages / day Shared across your entire organization. Resets at midnight UTC. Hourly company cap 200 messages / hour Throttles burst usage without blocking your team's day. Per-user daily cap 500 messages / user / day Prevents any single user from consuming the shared pool. ### Token limits at a glance Max prompt tokens4,096 / 8,000 Max completion tokens2,048 / 8,000 Daily token cap (company)1.4M / 2M Usage Allocation ## See exactly where your AI budget goes. Live counters, daily totals, and breakdowns by user, provider, and model - so you always know how AI is being used across your workspace. ### Today's usage Auto-refreshes every 30 seconds Messages used1,504 Tokens used1.4M Remaining496 Top users today Ravi M.312 msgs · 42% Priya K.198 msgs · 27% Meera S.114 msgs · 15% #### By provider & model See which AI model is handling the most requests. Switch or cap individual models without disrupting others. #### Daily totals & trends Filter by date range, user, or provider. Spot usage spikes before they hit your token cap. #### Per-session breakdown Drill into individual sessions - input tokens, output tokens, and model used per conversation. For developers ## Built open. Built to extend. Orangescrum AI is [MCP-native](/mcp) at its core. Plug in any model, embed the chat anywhere, or build entirely new workflows on top of it. #### Any AI model Use OpenAI, Anthropic, Gemini, or any provider. Configure via the admin panel - no code required. #### MCP protocol access Full read and write access via [MCP](/mcp) - tools automatically listed, callable by the model with confirmation gates for writes. #### Embed anywhere Drop it into your own tools, internal dashboards, or Slack integrations. The API is yours to build on. MCP server config // Connect Orangescrum to any AI model { "mcp\_servers": \[ { "type": "url", "name": "orangescrum", "url": "https://v4-api.orangescrum.com/mcp/partner" } \], "model": "your-model-here", // ↑ gpt-5-mini, claude-\*, gemini... "permissions": { "reads": true, "writes": "confirm" } } Write operations require explicit user confirmation - the AI never acts on your data silently. FAQ ## Frequently asked questions Quick answers about Orangescrum AI Chat - models, security, and setup. ### What is Orangescrum AI Chat? + Orangescrum AI Chat is a conversational interface that connects to your Orangescrum workspace through the [Model Context Protocol (MCP)](/mcp). You can ask plain-English questions about projects, [tasks](/task-management), sprints, and time logs, or take actions like creating tasks, [logging time](/time-tracking), and flagging blockers - all without leaving the chat. It uses your real, live project data; it doesn't guess or hallucinate. ### Which AI models can I use? + Any [MCP-compatible](/mcp) model. You can configure OpenAI (GPT-5, GPT-4), Anthropic (Claude), Google (Gemini), or any other provider via the admin panel - no code required. Switch models at any time. Bring-your-own-key is supported on [Enterprise plans](/contact-sales). ### What can the AI do besides answer questions? + Beyond read-only queries, the AI can [create and update tasks](/task-management), [log time](/time-tracking), change task statuses, assign work, flag blockers, generate [standup reports](/project-reports-analytics), and produce overdue summaries. Every write operation requires an explicit confirmation gate so the AI never modifies your data silently. ### Is my data sent to a third-party AI provider? + Only the context needed to answer a specific question is forwarded to the AI provider you configured. On Enterprise plans you can route through your own LLM endpoint or use a [self-hosted model](/self-hosted) so no project data leaves your network. All [MCP](/mcp) calls authenticate with your Orangescrum [partner API key](/developer/api) and run over TLS. ### Do I need a separate plan to use Orangescrum AI Chat? + Orangescrum AI Chat is included with your existing [Orangescrum plan](/pricing) during the early access period. It runs against the same [partner API key](/developer/api) you already have. [Contact your account manager](/contact-sales) for early access or to discuss Enterprise model bring-your-own-key configurations. ### How is this different from the Orangescrum MCP server? + [Orangescrum MCP](/mcp) is the protocol layer - the open standard that lets external AI clients (Claude Desktop, Cursor, Copilot) connect to your workspace. Orangescrum AI Chat is the in-product chat interface built on top of that same MCP layer, with confirmation gates, role-based access, and a UI optimized for project conversations. You can use either, both, or build your own client. Get started today ## Stop reporting. Start asking. Connect your Orangescrum workspace and get answers in under 5 minutes. No configuration, no training data, no setup friction. [Get early access - It's free](/sign-up)[Talk to sales](/contact-sales) No credit card required · Works with your existing Orangescrum plan