Jira Data Center 10 MCP Server (TypeScript)

A read-only Model Context Protocol (MCP) server for deep introspection of Jira Data Center 10 — 75 tools covering projects, workflows, schemes, fields, automation, Assets (Insight), and more.

Read-only — this server cannot modify any Jira configuration.

Why

Jira DC administration is config archaeology: the answer to "why did this happen?" hides somewhere in a chain of workflow → post-functions → automation rules → permission/field/screen schemes, and the admin UI shows that chain one screen at a time. This server hands the whole picture to an AI assistant in a single conversation — it can trace a project's full scheme chain, explain why an automation rule fired (or didn't), find every screen a custom field sits on, resolve who effectively holds a permission, or grep the server log, without SSH access and without any ability to change anything.

Most data comes from Jira's native REST API. The parts Jira only exposes through its Java API — listeners, scheduled services, effective permissions, transition rule internals, server logs — are unlocked by a companion set of ScriptRunner REST endpoints (a single GET-only Groovy file, see scriptrunner-endpoints/).

Related MCP server: Jira & Confluence MCP Server

Example prompts

• "Why didn't the automation rule 'Notify on escalation' fire for PROJ-123 yesterday?"

• "Compare the workflows of projects A and B and list every difference in transitions and post-functions."

• "Which screens, projects and issue types still use the custom field 'Severity'? Can I delete it?"

• "Who effectively has 'Delete Issues' permission in PROJ, and through which groups or roles?"

• "Grep atlassian-jira.log for indexing errors around 14:00 and summarize the stack traces."

Prerequisites

• Node.js 20 LTS or newer

• Jira Data Center 10.x

• A Jira PAT with Jira Administrator permission (for full config reads)

Building

npm install
npm run build

MCP client configuration

See .mcp.json.example for an example configuration file.

Response size

MCP clients cap the size of a single tool result (Claude Code drops results over ~25k tokens), and a large Jira has thousands of fields, rules, and workflows. So tool output is compact JSON, and instance-wide listings are paginated: they take offset/limit plus filters such as name_contains or project_key, and return { total, offset, returned, nextOffset, items }. A result that still exceeds JIRA_MCP_MAX_RESPONSE_CHARS (default 60000; 0 disables the check) is replaced by an error that tells the model how to narrow the call.

Self-test

npm run selftest exercises all 75 tools against the live Jira instance and prints a coverage report — one line per tool variant, then a summary. It auto-discovers the IDs/keys parameterised tools need (project keys, scheme IDs, workflow names, ...) from the list_*/dump_* tools, so no manual setup is needed. Credentials are read from the environment, falling back to the jira-dc server's env block in .mcp.json.

npm run selftest                          # all 75 tools
npm run selftest -- --only get_workflow   # one tool (deps auto-included)
npm run selftest -- --skip find_field_usage      # exclude slow tools

Each tool's raw JSON output is written to selftest-output/<tool>.json, with a full machine-readable report at selftest-output/_report.json. Flags: --only, --skip, --out-dir PATH, --json PATH.

Tools

All 75 tools are read-only against Jira (one, dump_script_registry, also writes its export bundle to a local directory). Tools marked † require the companion ScriptRunner Groovy endpoints (see scriptrunner-endpoints/).

Projects

• list_projects — all projects with key, name, type, lead

• get_project_config — a project's full scheme chain and issue types

• get_project_role_members — roles and their members/actors for a project

• get_project_components — components with leads and descriptions

• get_project_versions — versions with release status and dates

• list_project_categories — all project categories

Workflows

• list_active_workflows — active workflows (excludes backups and copies)

• list_all_workflows — all workflows including backups and deprecated

• get_workflow † — one workflow: statuses with category and meta properties, each transition once with its source statuses, screen, meta, and rule counts; with transition_ids (or "all") the full rules — condition tree, validators, pre/post-functions in execution order with arguments and decoded ScriptRunner scripts

• search_workflow_rules † — find a text (field id, class, user, group, jira.permission.* key, script fragment) across the rules and meta of all workflows

• list_workflow_schemes — all workflow schemes with issue-type mappings

• get_workflow_scheme — one workflow scheme by ID

Screens

• list_screens — all screens with IDs and names

• get_screen_tabs_and_fields — a screen's tabs and ordered fields

• list_screen_schemes — screen schemes with operation mappings; paginated

• get_screen_scheme — one screen scheme with screens and tabs

• list_issue_type_screen_schemes † — issue-type-to-screen-scheme mappings and projects; paginated, filter by project

• get_issue_type_screen_scheme — one issue type screen scheme

Fields

• list_fields — system and custom fields with types; paginated, filter by name

• list_custom_fields_usage — custom fields with usage stats (issues, projects, screens); paginated

• get_field_configuration — field config items: required, hidden, renderer; paginated, filter by field or required/hidden

• get_field_configuration_scheme — maps issue types to field configurations

• find_field_usage — where a field appears across all screens

• get_createmeta_fields — fields on the create screen for a project + issue type

• get_field_contexts — custom field contexts: project and issue-type scope

Schemes

• get_permission_scheme — full permission scheme with all grants

• list_permission_schemes — all permission schemes with grant counts

• get_notification_scheme — event-to-notification mappings

• list_notification_schemes — all notification schemes

• get_issue_type_scheme — available issue types and the default

• get_issue_security_scheme — issue security scheme with security levels

• get_priority_scheme — priority scheme (Jira DC 10)

Automation

• list_automation_rules — Automation for Jira (A4J) rules from the cache; paginated, filter by project or name

• get_automation_rule_detail — full rule: trigger, conditions, actions, smart values

• get_automation_audit_log — recent executions across all rules

• get_automation_rule_audit_log — execution history for one rule

• get_automation_audit_item — detail for a single audit entry

• refresh_automation_cache — force an automation cache refresh

Boards

• list_boards — all agile boards (Scrum/Kanban)

• get_board_configuration — columns, WIP limits, estimation, ranking, backing filter

Service Desk

• list_service_desks — all JSM service desks with project associations

• get_service_desk_queues — queues for a JSM service desk

• get_sla_config — SLA metrics (start/pause/stop conditions, goals), calendars, and config errors for a JSM project

Assets (Insight)

Read-only introspection of Atlassian Assets (formerly Insight) on Data Center. Uses the /rest/insight/1.0 API on the Jira host with IQL; override the base path with the optional ASSETS_API_BASE env var (default /rest/insight/1.0). The PAT user needs read access to the relevant Assets object schemas.

• list_object_schemas — all Assets object schemas with key, status, object/type counts

• get_object_schema — one object schema's metadata by ID

• list_object_types — object types in a schema (flat list, or hierarchical tree)

• get_object_type — one object type's metadata (parent, position, abstract, icon)

• get_object_type_attributes — attribute definitions for an object type (type, references, cardinality, required)

• get_schema_attributes — all attribute definitions across a schema

• list_object_statuses — Assets status types and categories (global or schema-scoped)

• search_objects_iql — search objects with IQL (Insight Query Language); first 25 matches by default, plus the total count

• get_object — one object by ID with its attribute values

• get_object_connected_tickets — Jira issues connected to an object

• dump_assets_schema — full structural snapshot: schema, statuses, and all object types with attributes

Filters & Dashboards

• list_filters — favourite/shared JQL filters

• list_dashboards — dashboards with view URL; paginated, filter by name

Analysis

• analyze_project_config_chain — resolve a project's scheme chain, report inconsistencies

• search_config — free-text search across all config entities

Issues

• get_issue — a Jira issue by key

• get_issue_changelog — an issue's edit history

Users

• get_user — user details by key, username, or ID

• find_users — search users by name or email

• get_user_groups — groups a user belongs to

• get_group_members — members of a group

Administration († ScriptRunner-backed)

• list_listeners † — registered event listeners

• list_scheduled_services † — scheduled services with cron schedules

• list_application_links † — links to Confluence, Bitbucket, Bamboo, etc.

• get_effective_permissions † — effective project permissions via groups, roles, grants

• dump_plugin_inventory — installed apps with version, enabled state, and license (user-installed by default)

Server logs († ScriptRunner-backed, System Administrators only)

• list_server_log_files † — log files in the Jira and Tomcat log directories with size and mtime; paginated, filter by name

• tail_server_log † — last N lines of a log file (default atlassian-jira.log)

• grep_server_log † — regex search over a log file and its rotations, with before/after context; matches return in chronological order

Dump

• dump_global_config — issue types, statuses, resolutions, priorities, link types, field counts

• dump_automation_rules — full A4J rules from the cache; paginated, filter by project or name

• dump_script_registry — full ScriptRunner inventory (listeners, REST endpoints, jobs, behaviours, fragments, script fields, resources, workflow functions, script-root .groovy files, instance metadata, Output.csv) reproducing ScriptRunner's "Export all scripts" bundle. Writes the bundle to a local directory (output_dir); reads from Jira are read-only.

License

MIT © sergeyopypey, Ivnrv, peppingdore