What Can a Jira AI Agent Change? A 97-Action Audit

What Can a Jira AI Agent Change? A 97-Action Audit

Steve Wall

A small mint reader passes through a cobalt permission boundary while a broad coral action field remains outside.

What Can a Jira AI Agent Change? A 97-Action Audit

A Jira-connected AI agent should not inherit every available Jira operation just because its account can open a project. In a bounded review of Aident Loadout's staging catalog on September 12, 2026, the current Jira integration contract contained 97 public Actions: 68 had retrieval or analysis-oriented names, while 29 had names that clearly add, create, edit, move, transition, notify, or delete Jira state.

That split is useful for discovery, but it is not a permission policy. One get Action can update the authenticated user's Recently viewed list, while one create-named Action only retrieves JQL autocomplete metadata. All 97 reviewed Actions also shared the wrapper's conservative operationType: write label and risk level 2. Authorize the exact Action, input, data scope, and consequence instead of trusting the integration name or the first verb.

For a safe first Jira test, allow only two jobs:

  1. list at most five viewable projects so a person can select the correct project; and

  2. return at most five recently updated issues from that approved project, limited to summary, status, and update time.

Do not authorize issue creation, edits, transitions, comments, notifications, attachments, worklogs, user or group changes, or deletion as part of that reader test.

If that is the job you need, inspect the current Jira integration in Aident Loadout before connecting an account. The useful next step is to review the live Actions and schemas, then preflight one bounded reader; it is not to enable the entire surface.

The 97-Action Audit Result

Observed name group

Count

What the name suggests

Policy treatment

get

50

Retrieve one object or reference set

Inspect fields, expansions, and input side effects before allowing

list, search, find, fetch, check, parse, and analysis

18

Discover or analyze Jira data

Bound project, query, returned fields, dates, and rows

create, add, delete, remove, assign, and other mutations

29

Change Jira or notify another person

Deny by default; approve one exact consequence and read-back at a time

Total public Jira Actions in the reviewed schema

97

A mixed read, write, and admin surface

Never approve from the connection or name alone

Actions labeled operationType: write by the reviewed wrapper

97

A conservative platform risk boundary

Use it as a gate, not as a taxonomy of the provider operation

The 68 retrieval and analysis-oriented names comprise 50 get, seven list, four search, two find, and one each beginning with analyse, check, evaluate, fetch, or parse. The 29 mutation-oriented names comprise eight create, six add, five delete, three remove, and one each beginning with assign, bulk, edit, move, send, transition, or update.

This count came from the exact Action enum behind the current public project-list contract, deduplicated and cross-checked against an integration-scoped capability search. It classifies names, not HTTP methods or business impact.

Names misclassify risk in both directions

The reviewed get issue schema accepts update_history. Its description says enabling that input adds the issue to the authenticated user's Recently viewed list. The issue content is read, but the viewing history changes.

The inverse also occurs. The reviewed create JQL autocomplete data Action says it retrieves field names, functions, and reserved words for query construction. Its create prefix sounds mutating even though the documented job is reference-data retrieval.

These two contracts make the durable rule concrete: review the current schema and consequence, not only the verb in the Action name.

What the Jira Surface Can Reach

The 97 Actions cross several responsibility boundaries.

Projects, boards, sprints, and versions

Reader-style Actions can enumerate projects, boards, sprints, project roles, components, versions, dashboards, and filters. The result may still reveal private project names, project leads, team structure, saved JQL, or planning data. A read-only role therefore needs a project boundary and a returned-field boundary.

Write-style Actions can create projects, boards, sprints, and versions or move issues into sprints. Those operations can alter planning, reporting, and the work visible to a team even when no issue description changes.

Issues, comments, attachments, and worklogs

The catalog can search and retrieve issues, comments, attachments, remote links, watchers, votes, transitions, and worklogs. These results may contain customer reports, internal discussions, personal information, filenames, time records, or incident details.

Other Actions can create, bulk-create, assign, edit, transition, link, or delete issues; add or update comments; send issue notifications; and add or delete attachments and worklogs. Each operation has a different audience and recovery cost. Permission to summarize five issues does not imply permission to comment on one.

Users, groups, roles, and permissions

The surface also includes users, groups, project roles, global and project permissions, and service-desk field metadata. Several Actions can add or remove users from groups and project roles. Those are administration consequences, not ordinary issue-management steps.

Atlassian's current Rovo MCP controls make a similar distinction. Its official server groups Jira tools under read, write, search, delete, and manage permissions, with delete and manage requiring admin enablement. Existing Jira user permissions still apply, while organization controls can narrow MCP access further. These layers help, but an agent still needs a task-specific allowlist.

Use a Jira Permission Ladder

Start with the smallest role that completes one useful job.

Role

Allow

Deny

Connection verifier

Account status, Vault status, capability discovery, and current schemas

Every Jira provider execution

Project reader

Five viewable projects with no expansions

Issues, comments, people, saved filters, writes, and admin data

Issue analyst

Five issues from one approved project, seven-day window, three returned fields

Descriptions, comments, attachments, worklogs, watchers, cross-project search, and every write

Reviewed issue operator

One exact approved mutation against one issue, followed by the smallest useful read-back

Any second mutation, expanded issue scope, notification, or user and group change

Jira administrator

Nothing by default

Projects, groups, roles, permission schemes, users, destructive Actions, and bulk changes without a separate admin policy

This ladder separates authentication, data access, and state change. A connected Jira account proves that a credential is available. It does not prove that the agent may inspect every project the person can see or exercise every permission that account holds.

A Safe First Test in Aident Loadout

Begin with the installed public CLI and the current connection state:

aident account auth status
aident vault status --integrationId "jira_tools"

If the Jira connection is not ready, stop after discovery, schema inspection, and preflight. Do not treat a valid Action schema as proof of provider access.

Discover the exact current project-list Action by job:

aident capabilities search \
  --query "Jira list viewable projects read only" \
  --types '["action"]' \
  --scope '{"integrationId":"jira_tools"}' \
  --targetEnv staging

Copy the exact public name returned by discovery, inspect it, and preflight a five-row request:

PROJECT_LIST_ACTION="<PASTE_RETURNED_PROJECT_LIST_ACTION>"

aident capabilities get \
  --name "$PROJECT_LIST_ACTION"

aident capabilities preflight \
  --name "$PROJECT_LIST_ACTION" \
  --input '{
    "maxResults":5,
    "action":"view",
    "orderBy":"name",
    "startAt":0
  }'

That exact placeholder-free input passed schema validation and returned a free Aident quote during this audit. It was not executed because the reviewed Vault had no connected Jira account. Avoid optional expansions until a person has reviewed what project metadata is necessary.

After a person selects one project key, discover and inspect the issue-search Action. Limit the response before execution:

ISSUE_SEARCH_ACTION="<PASTE_RETURNED_ISSUE_SEARCH_ACTION>"

aident capabilities get \
  --name "$ISSUE_SEARCH_ACTION"

aident capabilities preflight \
  --name "$ISSUE_SEARCH_ACTION" \
  --input '{
    "project_key":"<APPROVED_PROJECT_KEY>",
    "updated_after":"-7d",
    "fields":["summary","status","updated"],
    "max_results":5
  }'

This placeholder input also passed schema validation and returned a free quote. Replace the project key only after approval, then re-run preflight. Do not add description, comments, assignee, reporter, attachments, or custom fields unless the reader job requires them.

Success for the first live test is narrow: the agent returns no more than five issues from the selected project, each with only the approved fields, and makes no Jira change.

What Changes With Atlassian's MCP Server

Atlassian maintains an official MCP server and documents OAuth 2.1 and API-token authentication. Its current permission groups split Jira access into read_jira, write_jira, search_jira, and admin-enabled delete_jira and manage_jira. Atlassian also documents organization controls that can block MCP access by organization, site, container, or data classification even when the underlying user can open that content directly.

Those controls are useful enforcement layers. They do not replace the workflow's exact allowlist. A broad read_jira role can still expose more projects or issue fields than one analysis needs; write_jira can cover multiple consequences; and a user's existing permissions may be much broader than the temporary agent job.

Use the direct Atlassian route when Jira and Confluence are the entire task and the official permission groups match the operating policy. Use Aident Loadout when Jira is one reviewed step in a cross-service workflow and you need one discovery, Vault, preflight, and Audit pattern across the services. In either route, require human confirmation for high-impact or destructive actions and inspect the resulting Jira and audit state.

Record Six Things Before a Jira Write

Before an agent changes Jira, record:

  1. Identity: the Atlassian site, authenticated account, and selected project;

  2. Action: the exact current Action or MCP tool;

  3. Target: the issue, board, sprint, version, user, group, or project being changed;

  4. Input: every field, transition, comment, attachment, recipient, and notification option;

  5. Consequence: what becomes visible, changes workflow state, notifies people, or is difficult to recover; and

  6. Read-back: the smallest independent Jira read that proves only the approved change occurred.

For creation or bulk operations, add duplicate prevention. For edits, preserve the original values needed to review or reverse the change. For notifications and comments, preview the audience and text. One successful write does not grant standing permission for another issue or operation.

Common Jira Agent Permission Mistakes

Letting the person's Jira role become the agent's role

The agent operates through the person's permissions, but it should receive a narrower job-specific allowlist. A project administrator using an assistant to summarize five issues should not implicitly delegate project administration.

Treating all get Actions as side-effect free

The current issue-read schema can update Recently viewed history, and a read may expose sensitive fields or large expansions. Set optional side-effect inputs explicitly and request only necessary fields.

Treating all create names as writes

The JQL autocomplete Action retrieves reference data despite its prefix. Name-based blocking is a conservative starting point, not a substitute for schema review.

Approving search without a project or field ceiling

JQL and structured search can cross projects and return descriptions, people, or custom fields. Bind the project, dates, fields, and row limit before execution.

Using a free preflight as a safety verdict

Both bounded requests returned $0.00 Aident quotes. Free Actions can still expose private data or change provider state. Cost approval and consequence approval are separate controls.

Combining analysis and mutation in one autonomous step

An issue summary may recommend a transition, comment, assignment, or notification. Return that recommendation for review. Do not let evidence collection silently become a team-visible change.

Repeat the Audit When the Contract Changes

Recount and review the exact Action schemas when Atlassian changes its MCP permission model, Jira API behavior, or agent controls; when Aident changes the Jira Action version or wrapper metadata; or when the workflow's role expands.

This September 12 review is a contract audit, not a security certification or live-account test. It did not read private Jira data, execute a Jira Action, compare direct-MCP latency, test provider quotas, or verify all 97 operations against a real permission scheme. It establishes a narrower conclusion: the current surface is broad, names do not reliably encode consequences, and a bounded reader should be authorized separately from issue, workflow, notification, and administrative operators.

Inspect One Jira Reader Before Enabling Writes

Follow https://aident.ai/SETUP.md

Set up Aident Loadout and inspect one Jira reader

Use the risk-versus-spend approval workflow to separate a free quote from permission to touch Jira. Review Aident Loadout credits, BYOK, and Action pricing before any metered step. For the general discovery, connection, preflight, and execution loop, continue with How to Use Aident Loadout.

Sources

Refresh this audit when Atlassian changes its MCP authentication, permission groups, organization controls, Jira agent behavior, or provider APIs, or when Aident changes the Jira Action set, schemas, risk metadata, connection flow, pricing, or approval behavior.

About the author

Illustrated editorial avatar of Steve Wall

Steve Wall

Steve Wall is an editorial pen name used by Aident's workflow-research team. This column covers automation platform selection, adoption tradeoffs, and measuring workflow outcomes. Comparisons make their criteria, sources, assumptions, and limitations explicit so readers can evaluate the evidence for their own situation.

Home

Home

Home

Integrations

Integrations

Integrations

Vault

Vault

Vault

Audit

Audit

Audit

Arana Grande

Arana Grande

Arana Grande

Free

Free

Free

30-day audit summary

30-day audit summary

30-day audit summary

Daily action-call volume and the latest receipts from the Loadout audit trail.

Daily action-call volume and the latest receipts from the Loadout audit trail.

Daily action-call volume and the latest receipts from the Loadout audit trail.

View Audit

View Audit

View Audit

Loadout usage

Loadout usage

Loadout usage

617 action calls in the last 30 days

617 action calls in the last 30 days

617 action calls in the last 30 days

May 19 - Jun 17

May 19 - Jun 17

May 19 - Jun 17

10 active days

10 active days

10 active days

Less

Less

Less

More

More

More

Recent activity

Recent activity

Recent activity

Latest action-call receipts from connected agents

Latest action-call receipts from connected agents

Latest action-call receipts from connected agents

Apr 23, 09:23 AM

Apr 23, 09:23 AM

Apr 23, 09:23 AM

Shopify

Shopify

Shopify

Creates Or Updates An Asset For A Theme

Creates Or Updates An Asset For A Theme

Creates Or Updates An Asset For A Theme

Success

Success

Success

Apr 23, 09:21 AM

Apr 23, 09:21 AM

Apr 23, 09:21 AM

Shopify

Shopify

Shopify

Update Products Param Product Id

Update Products Param Product Id

Update Products Param Product Id

Success

Success

Success

Apr 23, 08:53 AM

Apr 23, 08:53 AM

Apr 23, 08:53 AM

Shopify

Shopify

Shopify

Update Products Param Product Id

Update Products Param Product Id

Update Products Param Product Id

Failed

Failed

Failed

Apr 22, 22:13 PM

Apr 22, 22:13 PM

Apr 22, 22:13 PM

Shopify

Shopify

Shopify

Create Product Image

Create Product Image

Create Product Image

Success

Success

Success

Apr 22, 22:12 PM

Apr 22, 22:12 PM

Apr 22, 22:12 PM

Shopify

Shopify

Shopify

Create Product Image

Create Product Image

Create Product Image

Success

Success

Success

Connected integration coverage

Connected integration coverage

Connected integration coverage

162

162

162

of 753 accessible connected

of 753 accessible connected

of 753 accessible connected

Callable actions

Callable actions

Callable actions

1,126

1,126

1,126

Vault credentials

Vault credentials

Vault credentials

8

8

8

Explore what's possible

Explore what's possible

Explore what's possible

See all Integrations

See all Integrations

See all Integrations

Google Ads

Google Ads

Google Ads

All available Goolge Ads tools via...

All available Goolge Ads tools via...

All available Goolge Ads tools via...

X (twitter)

X (twitter)

X (twitter)

All available X tools via...

All available X tools via...

All available X tools via...

Github

Github

Github

All available Github tools via...

All available Github tools via...

All available Github tools via...

Notion

Notion

Notion

All available Notion tools via...

All available Notion tools via...

All available Notion tools via...

Slack

Slack

Slack

All available Slack tools via...

All available Slack tools via...

All available Slack tools via...

Firecrawl

Firecrawl

Firecrawl

All available Firecrawl tools via...

All available Firecrawl tools via...

All available Firecrawl tools via...

753 integrations are available for loadouts.

753 integrations are available for loadouts.

753 integrations are available for loadouts.

The one tool

for every tool

your agent needs.

Give any AI agent real capabilities in seconds. Connect 27,000+ tools once, skip the setup headache, and let your agents execute.

Try Aident Loadout

Empower your Codex or OpenClaws to get real jobs done. Connect 27,000+ tools in one prompt, and let your agents deliver real results.

Try Aident Loadout

Empower your Codex or OpenClaws to get real jobs done. Connect 27,000+ tools in one prompt, and let your agents deliver real results.

Try Aident Loadout

Empower your Codex or OpenClaws to get real jobs done. Connect 27,000+ tools in one prompt, and let your agents deliver real results.