atlassian-browser-mcp

MCP server that wraps the upstream mcp-atlassian toolset with browser-cookie authentication. By default it reuses cookies from an existing Firefox profile, so you can sign in through your normal browser and avoid API tokens. A Playwright-based manual login mode is still available when explicitly enabled.

How it works

1. Sign in to Jira and Confluence in Firefox using your normal SSO/MFA flow
2. On startup, the server loads matching Firefox cookies with browser-cookie3
3. All MCP tool calls use those cookies via a custom requests.Session subclass
4. If an API response looks like an SSO redirect, the session reloads Firefox cookies and retries once

The server monkey-patches JiraClient and ConfluenceClient constructors in mcp-atlassian to inject the browser-backed session, giving full parity with the upstream tool surface (73 Atlassian tools + 1 atlassian_login helper = 74 total).

Files

File	Purpose
atlassian_browser_mcp_full.py	Entrypoint. Patches upstream clients, registers atlassian_login tool, runs the MCP server
atlassian_browser_auth.py	Shared auth: BrowserCookieSession, cookie providers, SSO detection
run-atlassian-browser-mcp.sh	Launcher: creates venv, installs deps via uv, runs compatibility check, starts server
pyproject.toml	Dependency pins

Usage

First, open Firefox and sign in to Jira and Confluence. Then run:

export JIRA_URL="https://jira.example.com"
export CONFLUENCE_URL="https://confluence.example.com"
./run-atlassian-browser-mcp.sh

Or configure as an MCP server in your editor (Cursor, Claude Code, etc.) pointing to the launcher script.

This server exposes Atlassian actions as MCP tools, not MCP resources. If your client reports resources/list failed or resources/templates/list failed, call the Jira/Confluence tools directly instead of resource discovery.

Tool Reference

Tool names below are the exposed MCP names in Codex. The wrapper prefixes Jira tools with jira_ and Confluence tools with confluence_.

Jira tools

Tool	Used for
jira_get_user_profile	Retrieve profile information for a specific Jira user.
jira_get_issue_watchers	Get the list of watchers for a Jira issue.
jira_add_watcher	Add a user as a watcher to a Jira issue.
jira_remove_watcher	Remove a user from watching a Jira issue.
jira_get_issue	Get details of a specific Jira issue including its Epic links and relationship information.
jira_search	Search Jira issues using JQL (Jira Query Language).
jira_search_fields	Search Jira fields by keyword with fuzzy match.
jira_get_field_options	Get allowed option values for a custom field.
jira_get_project_issues	Get all issues for a specific Jira project.
jira_get_transitions	Get available status transitions for a Jira issue.
jira_get_worklog	Get worklog entries for a Jira issue.
jira_download_attachments	Download attachments from a Jira issue.
jira_get_issue_images	Get all images attached to a Jira issue as inline image content.
jira_get_agile_boards	Get Jira agile boards by name, project key, or type.
jira_get_board_issues	Get all issues linked to a specific board filtered by JQL.
jira_get_sprints_from_board	Get Jira sprints from board by state.
jira_get_sprint_issues	Get Jira issues from sprint.
jira_get_link_types	Get all available issue link types.
jira_create_issue	Create a new Jira issue with optional Epic link or parent for subtasks.
jira_batch_create_issues	Create multiple Jira issues in a batch.
jira_batch_get_changelogs	Get changelogs for multiple Jira issues (Cloud only).
jira_update_issue	Update an existing Jira issue including changing status, adding Epic links, updating fields, etc.
jira_delete_issue	Delete an existing Jira issue.
jira_add_comment	Add a comment to a Jira issue.
jira_edit_comment	Edit an existing comment on a Jira issue.
jira_add_worklog	Add a worklog entry to a Jira issue.
jira_link_to_epic	Link an existing issue to an epic.
jira_create_issue_link	Create a link between two Jira issues.
jira_create_remote_issue_link	Create a remote issue link (web link or Confluence link) for a Jira issue.
jira_remove_issue_link	Remove a link between two Jira issues.
jira_transition_issue	Transition a Jira issue to a new status.
jira_create_sprint	Create Jira sprint for a board.
jira_update_sprint	Update Jira sprint.
jira_add_issues_to_sprint	Add issues to a Jira sprint.
jira_get_project_versions	Get all fix versions for a specific Jira project.
jira_get_project_components	Get all components for a specific Jira project.
jira_get_all_projects	Get all Jira projects accessible to the current user.
jira_get_service_desk_for_project	Get the Jira Service Desk associated with a project key.
jira_get_service_desk_queues	Get queues for a Jira Service Desk.
jira_get_queue_issues	Get issues from a Jira Service Desk queue.
jira_create_version	Create a new fix version in a Jira project.
jira_batch_create_versions	Batch create multiple versions in a Jira project.
jira_get_issue_proforma_forms	Get all ProForma forms associated with a Jira issue.
jira_get_proforma_form_details	Get detailed information about a specific ProForma form.
jira_update_proforma_form_answers	Update form field answers using the Jira Forms REST API.
jira_get_issue_dates	Get date information and status transition history for a Jira issue.
jira_get_issue_sla	Calculate SLA metrics for a Jira issue.
jira_get_issue_development_info	Get development information (PRs, commits, branches) linked to a Jira issue.
jira_get_issues_development_info	Get development information for multiple Jira issues.

Confluence tools

Tool	Used for
confluence_search	Search Confluence content using simple terms or CQL.
confluence_get_page	Get content of a specific Confluence page by its ID, or by its title and space key.
confluence_get_page_children	Get child pages and folders of a specific Confluence page.
confluence_get_space_page_tree	Get page hierarchy for a Confluence space as a flat list.
confluence_get_comments	Get comments for a specific Confluence page.
confluence_get_labels	Get labels for Confluence content (pages, blog posts, or attachments).
confluence_add_label	Add label to Confluence content (pages, blog posts, or attachments).
confluence_create_page	Create a new Confluence page.
confluence_update_page	Update an existing Confluence page.
confluence_delete_page	Delete an existing Confluence page.
confluence_move_page	Move a Confluence page to a new parent or space.
confluence_add_comment	Add a comment to a Confluence page.
confluence_reply_to_comment	Reply to an existing comment thread on a Confluence page.
confluence_search_user	Search Confluence users using CQL (Cloud) or group member API (Server/DC).
confluence_get_page_history	Get a historical version of a specific Confluence page.
confluence_get_page_diff	Get a unified diff between two versions of a Confluence page.
confluence_get_page_views	Get view statistics for a Confluence page.
confluence_upload_attachment	Upload an attachment to Confluence content (page or blog post).
confluence_upload_attachments	Upload multiple attachments to Confluence content in a single operation.
confluence_get_attachments	List all attachments for a Confluence content item (page or blog post).
confluence_download_attachment	Download an attachment from Confluence as an embedded resource.
confluence_download_content_attachments	Download all attachments for a Confluence content item as embedded resources.
confluence_delete_attachment	Permanently delete an attachment from Confluence.
confluence_get_page_images	Get all images attached to a Confluence page as inline image content.

Helper

Tool	Used for
atlassian_login	Refresh browser-backed Atlassian authentication.

Minimize Codex token usage

If you use this server from Codex, reduce token usage by redacting the MCP tool list in Codex itself. Codex supports per-server tool allowlists and denylists in ~/.codex/config.toml via enabled_tools and disabled_tools; see the official Codex MCP docs and config reference.

For the lowest token usage, prefer enabled_tools. That prevents all non-listed tool schemas from being exposed to Codex for this MCP server.

Example: keep only basic Jira lookup tools available in Codex:

[mcp_servers.atlassian_browser]
command = "/Users/you/Projects/atlassian-browser-mcp/run-atlassian-browser-mcp.sh"
cwd = "/Users/you/Projects/atlassian-browser-mcp"
startup_timeout_sec = 30
tool_timeout_sec = 120
enabled_tools = [
  "jira_search",
  "jira_get_issue",
  "jira_download_attachments", 
  "jira_get_issue_images"
]

[mcp_servers.atlassian_browser.env]
JIRA_URL = "https://jira.example.com"
CONFLUENCE_URL = "https://confluence.example.com"

If you want to block only a few tools and keep the rest, use disabled_tools instead:

[mcp_servers.atlassian_browser]
command = "/Users/you/Projects/atlassian-browser-mcp/run-atlassian-browser-mcp.sh"
cwd = "/Users/you/Projects/atlassian-browser-mcp"
disabled_tools = [
  "confluence_search",
  "confluence_get_page",
]

Notes:

• enabled_tools is better than disabled_tools when your goal is token reduction.
• Redaction is client-side in Codex. Other MCP clients still see the full tool surface unless they apply their own filtering.
• Restart Codex after changing ~/.codex/config.toml.
• Add atlassian_login to enabled_tools only if you need the manual login helper exposed.

Use a Jira summarizer subagent in Codex

For even lower main-session token usage, keep Atlassian MCP tools out of the main Codex session and expose them only to a focused Jira summarizer subagent. Codex custom agents can live in ~/.codex/agents/ for personal use or .codex/agents/ for project-scoped use. Codex only spawns subagents when you explicitly ask it to.

Use gpt-5.4-mini with model_reasoning_effort = "medium" by default. Jira summarization is mostly retrieval, filtering, and compression. Use gpt-5.4 with model_reasoning_effort = "high" only when the Jira work is complex, spans multiple tickets, or needs careful risk synthesis. See the official Codex subagents docs, Codex MCP docs, Codex config reference, and model list.

Create ~/.codex/agents/jira-summarizer.toml:

name = "jira_summarizer"
description = "Reads Jira issues through atlassian-browser-mcp and returns compact summaries for the parent agent."
model = "gpt-5.4-mini"
model_reasoning_effort = "medium"
sandbox_mode = "read-only"

developer_instructions = """
Use Jira tools only to gather issue context.
Summarize for the parent agent, not for an end user.
When reading a specific issue, include comments because they often contain decisions, clarifications, and current blockers.
Check for attachments and download them when filenames, metadata, or the parent request suggest they may contain useful requirements, screenshots, logs, or reproduction details.
Prefer compact output: status, assignee, priority, key dates, requirements, blockers, comments, attachment findings, and engineering implications.
Do not make code changes.
Do not fetch unrelated issues unless the parent agent asks for linked or related work.
"""

[mcp_servers.atlassian_browser]
command = "/Users/you/Projects/atlassian-browser-mcp/run-atlassian-browser-mcp.sh"
cwd = "/Users/you/Projects/atlassian-browser-mcp"
startup_timeout_sec = 30
tool_timeout_sec = 120
enabled_tools = [
  "jira_search",
  "jira_get_issue",
  "jira_download_attachments",
  "jira_get_issue_images",
]

[mcp_servers.atlassian_browser.env]
JIRA_URL = "https://jira.example.com"
CONFLUENCE_URL = "https://confluence.example.com"

Then ask the main agent to delegate Jira reading:

Use jira_summarizer to read BIZ-20528, include comments, attachments, status, assignee, and acceptance details if available, and return a compact summary with blockers and engineering implications.

Firefox profile discovery is automatic. If Firefox has multiple profiles and the wrong one is selected, point directly at the cookie database:

export FIREFOX_COOKIE_FILE="$HOME/Library/Application Support/Firefox/Profiles/xxxx.default-release/cookies.sqlite"
./run-atlassian-browser-mcp.sh

To use the previous Playwright/Chromium login flow instead:

export COOKIES_PROVIDER=playwright
./run-atlassian-browser-mcp.sh

Environment variables

Variable	Default	Description
JIRA_URL	(required)	Jira base URL (e.g. https://jira.example.com)
CONFLUENCE_URL	(required)	Confluence base URL (e.g. https://confluence.example.com)
ATLASSIAN_BROWSER_AUTH_ENABLED	true	Enable browser auth (set false to fall back to token auth)
COOKIES_PROVIDER	firefox	Cookie provider. Use firefox for existing Firefox cookies or playwright for the legacy Chromium login flow
FIREFOX_COOKIE_FILE	(auto)	Optional path to a Firefox cookies.sqlite file when automatic discovery picks the wrong profile
HTTP_PROXY, HTTPS_PROXY, ALL_PROXY, NO_PROXY	(none)	Standard proxy environment variables honored by the underlying requests.Session; lowercase variants are also supported
ATLASSIAN_BROWSER_PROFILE_DIR	./.atlassian-browser-profile	Playwright mode only: persistent Chromium profile directory
ATLASSIAN_STORAGE_STATE	./.atlassian-browser-state.json	Playwright mode only: storage-state file
ATLASSIAN_LOGIN_TIMEOUT_SECONDS	300	Playwright mode only: seconds to wait for manual login
ATLASSIAN_USERNAME	(none)	Playwright mode only: prefill username on SSO page
ATLASSIAN_SSO_MARKERS	(auto)	Comma-separated URL/text markers for SSO redirect detection. Defaults cover Okta, ADFS, Azure AD, PingOne, Google SAML
TOOLSETS	all	Which upstream toolsets to enable

Requirements

• Python 3.12+
• uv (for dependency management)
• Firefox with an active Jira/Confluence session
• Chromium only when COOKIES_PROVIDER=playwright (installed automatically by Playwright)
• Network access to your Atlassian instance