Content
View differences
Updated by Dominic Bräunlein 3 months ago
# User Problem
## User
* Project managers who want to query, summarize, or report on projects naturally using AI tools. Administrators
* Product owners who want to automate repetitive project reporting.
* Project members who need quick summaries, comparisons, managers exploring AI or contextual insights from project data.
* Admins who want to safely connect OpenProject data with AI assistants. automation integrations
<br>
## Problem
* Users want to interact with OpenProject using natural language with need a standardized and secure way for external AI assistants agents or automation tools to generate project status reports, find work packages, or analyze progress.
* Currently, they must extract data manually, build API scripts, or rely access and act on nonstandard integrations. OpenProject data.
* There’s no consistent, secure, standard protocol Administrators want to let LLMs work with centrally manage and control these connections without exposing the full OpenProject data. API or custom credentials.
<br>
<br>
## Pain
* Manually copy project data for analysis Developers manually build integrations using the REST API
* Use API tokens in Each custom scripts which is a technical barrier implementation increases maintenance cost, security risk, and not maintainable. complexity.
* The lack of a unified protocol prevents reuse of connectors and compatibility with emerging AI models can’t access project data directly
<br> agent standards like the Model Context Protocol (MCP).
# Business Case
## Solution
* Implement a an MCP server for OpenProject that exposes key project data and actions through the Model Context Protocol (MCP) server within OpenProject, using the official Ruby MCP SDK. (JSON-RPC 2.0).
* Use HTTP Support stdin/stdout (local) or SSE/HTTP (remote) transport only (no WebSocket or streaming) mechanisms for simpler infrastructure compatibility with local and easier scaling. hosted agents.
* Reuse OpenProject’s existing OAuth2 authentication (with PKCE) Integrate with OAuth for secure access — compatible with both browser-based authentication and desktop MCP clients. authorization.
* Provide two search tools out of the box:
* searchProject: filters projects by standard attributes (name, identifier, status).
* searchWorkPackage → filters an initial toolset: list and read projects, work packages, and comments, and create work packages by standard attributes (subject, type, status, project, version, user).
* Expose resources derived from OpenProject APIv3:
* projects
* work\_packages
* versions
* users
* Add an via standardized MCP admin page under Administration -> AI Integration: methods.
* Allows renaming or localizing tool/resource names and descriptions.
* Enables or disables specific tools/resources for compliance or privacy.
<br>
<br>
## Out of Scope for the MVC
* Custom field support
* Full write access for all entities in OpenProject
* Permissions Permission to control MCP usage
* Dynamic tool discovery based on role permission
* WebSocket or Server-Sent Events (streaming).
* AI model invocation or prompt execution inside OpenProject.
* Complex dynamic prompt management or workflow automation
* Full import of APIv3 resources
* Granular field-level permissions for MCP exposure
## Differentiation
* One of the first project management platform with native MCP support, enabling direct AI assistant access to real project data.
* Fully integrated with Maintains OpenProject’s existing authentication security and permissions. permission model while exposing a universal interface.
* Admin customization layer to align Enables direct connection between OpenProject and AI vocabulary with organizational terminology (e.g., “features,” “tickets,” or “projectOne”).
* Simpler architecture (HTTP-only, no streaming) and easier deployment for on-premise customers.
<br> assistants without custom plugins.
## Next iteration
* Add prompt support for standardized templates (e.g., “Project Status Report,” “Sprint Summary,” “Risk Overview”).
* Extend _What is the MCP admin UI next solution that would allow us to manage prompt templates and output schemas.
release meaningful customer value quickly?_
* Add semantic search for projects and work packages. ...
* Optional support for streaming / real-time notifications.
* Expand resource catalog (meetings, comments, attachments).
<br>
# Launch and Growth
## Measures
* Number of successful MCP client connections. Successful connection and operation from an MCP-compatible client.
* Time saved on automated workflows vs. manual workflows.
* Positive feedback from AI and integration partners during pilot customers integrating with AI agents. testing.
<br>
## Messaging
_If you were to write a press release, how would you describe the value to customers?_
<figure class="table op-uc-figure_align-center op-uc-figure"><table class="op-uc-table"><tbody><tr class="op-uc-table--row"><th class="op-uc-table--cell op-uc-table--cell_head"><p class="op-uc-p">Headline</p></th><td class="op-uc-table--cell"><p class="op-uc-p">OpenProject connects directly with AI assistants through the Model Context Protocol</p></td></tr><tr class="op-uc-p"><br data-cke-filler="true"></p></td></tr><tr class="op-uc-table--row"><th class="op-uc-table--cell op-uc-table--cell_head"><p class="op-uc-p">First Paragraph</p></th><td class="op-uc-table--cell"><p class="op-uc-p">OpenProject now integrates seamlessly connects with AI assistants and automation tools via through the open Model Context Protocol (MCP). This allows users new server enables secure, standardized access to securely query, summarize, projects and report on real project data by using natural language, work packages, empowering teams to integrate intelligent workflows and assistants without complex custom scripts or manual exports.</p></td></tr><tr integrations.</p></td></tr><tr class="op-uc-table--row"><th class="op-uc-table--cell op-uc-table--cell_head"><p class="op-uc-p">Customer Quote</p></th><td class="op-uc-table--cell"><p class="op-uc-p">-</p></td></tr></tbody></table></figure> class="op-uc-p"><br data-cke-filler="true"></p></td></tr></tbody></table></figure>
##
# Further Resources
* [https://modelcontextprotocol.io/introduction](https://modelcontextprotocol.io/introduction)
* Example showing how to expose functionality of Ghidra (Reverse Engineering Tool) to an LLM and how it is used, along with some nice applications: https://www.youtube.com/watch?v=u2vQapLAW88
## User
* Project managers who want to query, summarize, or report on projects naturally using AI tools.
* Product owners who want to automate repetitive project reporting.
* Project members who need quick summaries, comparisons,
* Admins who want to safely connect OpenProject data with AI assistants.
<br>
## Problem
* Users want to interact with OpenProject using natural language with
* Currently, they must extract data manually, build API scripts, or rely
* There’s no consistent, secure, standard protocol
<br>
<br>
## Pain
* Manually copy project data for analysis
* Use API tokens in
*
<br>
# Business Case
## Solution
* Implement a
* Use HTTP
* Reuse OpenProject’s existing OAuth2 authentication (with PKCE)
* Provide two search tools out of the box:
* searchProject: filters projects by standard attributes (name, identifier, status).
* searchWorkPackage → filters
* Expose resources derived from OpenProject APIv3:
* projects
* work\_packages
* versions
* users
* Add an
* Allows renaming or localizing tool/resource names and descriptions.
* Enables or disables specific tools/resources for compliance or privacy.
<br>
<br>
## Out of Scope for the MVC
* Custom field support
* Full write access for all entities in OpenProject
* Permissions
* Dynamic tool discovery based on role permission
* WebSocket or Server-Sent Events (streaming).
* AI model invocation or prompt execution inside OpenProject.
* Complex dynamic prompt management or workflow automation
* Full import of APIv3 resources
* Granular field-level permissions for MCP exposure
## Differentiation
* One of the first project management platform with native MCP support, enabling direct AI assistant access to real project data.
* Fully integrated with
* Admin customization layer to align
* Simpler architecture (HTTP-only, no streaming) and easier deployment for on-premise customers.
<br>
## Next iteration
* Add prompt support for standardized templates (e.g., “Project Status Report,” “Sprint Summary,” “Risk Overview”).
* Extend
* Optional support for streaming / real-time notifications.
* Expand resource catalog (meetings, comments, attachments).
<br>
# Launch and Growth
## Measures
* Number of successful MCP client connections.
* Time saved on automated workflows vs. manual workflows.
* Positive feedback from
<br>
## Messaging
##
# Further Resources
* [https://modelcontextprotocol.io/introduction](https://modelcontextprotocol.io/introduction)
* Example showing how to expose functionality of Ghidra (Reverse Engineering Tool) to an LLM and how it is used, along with some nice applications: https://www.youtube.com/watch?v=u2vQapLAW88