- An open standard for connecting AI to tools.
- Secure, and backed by the tech giants.
- Already implemented by Atlassian for Jira and Confluence.
One of the main challenges of artificial intelligence in the enterprise is connecting it simply and securely to everyday collaboration tools such as Jira or Confluence, without creating technical debt or compromising data governance.
A standardised answer is starting to emerge: MCP, the Model Context Protocol. At BleuLemon, we believe it is an effective technology, one that balances innovation against the stability of your systems.
1. What is MCP, concretely?
MCP is the acronym for Model Context Protocol.
It is a specification published and maintained by the open-source organisation (modelcontextprotocol).
MCP describes a JSON-RPC dialogue between 3 roles:
- An AI interface (chatbot)
- A connector (MCP server)
- A third-party server (the source of data or of functions, for example a Jira/Confluence server)
The purpose of MCP is to standardise the way applications pass context and expose "tools" to an LLM. This extends the finite capabilities of a plain LLM by giving it read or write access to external resources, while preserving the scalability of your infrastructure.
In practice, MCP acts as a connector where each party brings:
- Resources: data (Jira tickets, web content, files, and so on)
- Tools: actions the model can execute (create a ticket, launch a build, and so on)
- Prompts: message or workflow templates
Anthropic, which initiated the standard, released MCP as open source at the end of November 2024 to “break down data silos” and speed up the creation of autonomous agents. This standardised approach to orchestrating AI agents answers a growing need for interoperability in the enterprise. The initiative is already backed by OpenAI, Google and Microsoft, and several dozen community servers now cover a variety of use cases.
Why the excitement?
- Interoperability: an MCP client (LLM chatbot + MCP) can plug into any server without adapting its code.
- Explicit security: every call is documented, signed and validated on the user side before execution, which meets data governance requirements.
- An ecosystem in full ferment: the community of users, data scientists and data engineers is taking hold of it, experimenting with it and imagining new applications every day.
2. Atlassian's official MCP Server
Announced in beta on 1 May 2025, Atlassian's Remote MCP Server is the first turnkey implementation that exposes Jira and Confluence Cloud to AI agents, starting with Anthropic's Claude.
Atlassian hosts the infrastructure on Cloudflare, with OAuth authentication and strict respect for existing Atlassian permissions.
Main features and limitations of the Atlassian MCP server
- summarise a ticket, create a page, or chain several actions in a single request.
Quotas aligned with the Atlassian Cloud plan (atlassian.com):
- 500 calls/hour free, whatever the plan
- 1,000 calls/hour on the Standard plan
- up to 10,000 calls/hour on the Premium/Enterprise plan (plus 20 calls per user)
Current limitations:
- Atlassian Cloud Jira and Confluence only, with Anthropic's Claude chatbot.
Advantages
Points to watch
Complete actions:Read and write are both possible (summarise a ticket, create a page, chain several actions in a single request)
Limited ecosystem:Compatible only with Atlassian Cloud (Jira + Confluence) and Anthropic's Claude chatbot
Generous quotas: - 500 calls/hour (free, all plans)
- 1,000 calls/hour (Standard)- Up to 10,000 calls/hour (Premium/Enterprise)
Security risks:Exposure to prompt-injection attacks, "rug pull" and tool redefinition (Tool Redefinition)
Secure infrastructure:Cloudflare hosting, OAuth authentication, existing Atlassian permissions respected
Beta status:Features and stability still being validated, changes possible
Ready for production:Quotas large enough to move from POC to daily enterprise use
Governance required:Systematic log auditing and version pinning of third-party servers recommended
On the security side, Atlassian sets out the risks specific to MCP clients and servers (prompt-injection, “rug pull” or Tool Redefinition, and so on) and recommends the principle of least privilege, systematic log auditing and version pinning of third-party servers (atlassian.com).
“MCP servers already expose gateways to code repositories, chat spaces and databases. Combined with issue tracking, they finally orchestrate the whole product flow, from commit to request, simply in natural language. To my mind this is the future of how software tools are used.
At BleuLemon, we are running more and more MCP tests, because we believe this protocol will redefine the way technical teams interact with their software stack."
Worth a look: Atlassian has published a short video demonstration of the Remote MCP Server in use:
(Atlassian Remote MCP Server video, 25 sec)
3. And now: test the first use cases for your MCP server
How much time do your teams lose every day juggling the Jira window, the Confluence window and the chat?
Early feedback from the teams testing the beta shows that the pairing of Claude + Remote MCP Server already opens the door to several concrete uses:
- Instant summaries of a Jira ticket or a Confluence page ahead of a meeting.
- Direct creation of tickets or pages from the conversation, without leaving the chatbot.
- Chained actions: generate 10 issues from a backlog, then publish the release notes page, then notify the team.
- Automatic enrichment of a ticket with context drawn from other sources the agent can reach.
These scenarios are only a foretaste: Atlassian already plans to extend the integration to other products and offers quotas generous enough to move from POC to daily use.
And what about Data Center environments?
Not everyone is on Atlassian's Cloud offering. In parallel we are working on open-source MCP servers to bring the same capabilities to Data Center (on-prem) instances, along with the added value that goes with them for our clients in that environment.
Stay tuned: our next article will detail this work and share our first findings from the field.
How can BleuLemon support you?
The idea appeals to you, but the project looks complex? BleuLemon is already helping its clients to:
- Assess how relevant MCP is in your own context and calculate the gains from automation.
- Build a POC to demonstrate the value to your teams while respecting your data governance constraints.
- Deploy and secure a solution across your whole organisation, with a controlled approach to orchestrating AI agents.
BleuLemon supports IT departments through their transformation with freshness and quickness of mind. We help organisations unlock the collective potential of their teams, opening the way to a stronger and more fulfilling way of organising work.