Axial API
This article and the rest of the API documentation in this section are written for a technical audience — integrators and developers connecting external systems to Tickiti. Familiarity with HTTP, REST, JSON and bearer-token authentication is assumed.
The Axial API lets a script or an AI assistant work with Axial, Tickiti’s AI assistant, as a member of staff does: read and steer Axial’s work on a ticket through everything the Axial panel offers, hold their own conversations with Axial outside any ticket, and keep the gap log. Axial’s settings, skill store, usage and reports are in the Axial settings API.
Every call goes through the panel’s own code as the token’s owner, so their Axial access on the ticket, the queue, and the rights an action asks for all apply exactly as they do in the panel. What a call writes is marked as coming through the API, with the token’s name. The endpoints need a plan that includes Axial and a token owned by a member of staff. All are POST, take and return JSON, and use the response envelope described in the API overview.
Abilities
The axial-chat family has four abilities, granted separately:
axial-chat:read— read the panel, the runs and the gap log. Anyaxial-chat:ability reads.axial-chat:write— talk to Axial: messages, Update suggested response, Respond again, adding a note, cancelling your own waiting request, and agent tasks.axial-chat:approve— answer a question card, approve or reject a proposed action, and act on a shared note’s card. Approving is acting: the action runs at once, as the token’s owner.axial-chat:manage— Rerun, choose the ticket’s skill, Turn back on, approve translations, apply or discard a skill change Axial drafted, and update the gap log.
A call is refused while someone else holds the ticket’s Axial panel, as the panel itself would refuse them; an API call takes no hold of its own.
Endpoints
| Method & path | Ability | Purpose |
|---|---|---|
| POST /api/v1/axial/chat | axial-chat:read | The panel as staff see it |
| POST /api/v1/axial/chat/runs | axial-chat:read | Each time Axial worked on the ticket |
| POST /api/v1/axial/chat/message | axial-chat:write | Write to Axial in the ticket’s chat |
| POST /api/v1/axial/chat/check-wait | axial-chat:write | Check now on work Axial is waiting for |
| POST /api/v1/axial/chat/redraft | axial-chat:write | Update suggested response |
| POST /api/v1/axial/chat/respond-again | axial-chat:write | Respond again after a reply was sent |
| POST /api/v1/axial/chat/note | axial-chat:write | Add an internal note to Axial’s conversation |
| POST /api/v1/axial/chat/queue/cancel | axial-chat:write | Take back your own waiting request |
| POST /api/v1/axial/chat/agent-task | axial-chat:write | Give Axial a task |
| POST /api/v1/axial/chat/agent-task/deep | axial-chat:write | Investigate in depth |
| POST /api/v1/axial/chat/agent-task/stop | axial-chat:write | Stop an agent task |
| POST /api/v1/axial/chat/card | axial-chat:approve | Answer or dismiss a question card |
| POST /api/v1/axial/chat/approval | axial-chat:approve | Approve or reject a proposed action |
| POST /api/v1/axial/chat/shared-note | axial-chat:approve | Act on a shared note’s card |
| POST /api/v1/axial/chat/rerun | axial-chat:manage | Rerun the latest customer message |
| POST /api/v1/axial/chat/skill | axial-chat:manage | Choose the ticket’s skill |
| POST /api/v1/axial/chat/turn-on | axial-chat:manage | Turn back on |
| POST /api/v1/axial/chat/translations/approve | axial-chat:manage | Approve a run’s stock-response translations |
| POST /api/v1/axial/chat/skill-proposal | axial-chat:manage | Apply or discard a skill change Axial drafted |
| POST /api/v1/axial/conversations | axial-chat:read | Your conversations with Axial |
| POST /api/v1/axial/conversations/show | axial-chat:read | One conversation |
| POST /api/v1/axial/conversations/create | axial-chat:write | Start a conversation |
| POST /api/v1/axial/conversations/message | axial-chat:write | Continue a conversation |
| POST /api/v1/axial/conversations/stop | axial-chat:write | Stop Axial answering |
| POST /api/v1/axial/conversations/update | axial-chat:write | Rename or delete a conversation |
| POST /api/v1/axial/gaps | axial-chat:read | The gap log |
| POST /api/v1/axial/gaps/update | axial-chat:manage | Change or merge a gap |
A ticket’s Axial panel
Every axial/chat endpoint takes the ticket as ticket_number or ticket_id.
Reading. axial/chat returns the conversation with Axial, the latest suggested response with its note to staff, every question card with its answer and what its actions did, and anything waiting — including requests queued behind Axial’s current work, each with an id. since returns only turns after a given one; include adds Axial’s working (tools) and earlier versions of the suggested response (drafts). axial/chat/runs lists each time Axial worked on the ticket, newest first, with the skill and models used, status, confidence and cost; limit caps it.
Talking to Axial. chat/message takes message and writes it in the ticket’s chat; nothing reaches the customer. chat/redraft writes the suggested response again over the whole conversation, and chat/respond-again writes another one to the same customer message after a reply has gone. Both wait their turn while Axial is busy, as in the panel. chat/note takes the response_id of one of the ticket’s internal notes and adds it to Axial’s conversation (notes are left out unless someone adds them). chat/queue/cancel takes the id of one of your own waiting requests. chat/check-wait takes a run_id or task_id and checks whether long-running work Axial is waiting for has finished.
Cards and approvals. chat/card takes card_id, decision (answer or dismiss), the answers per question and an optional note. chat/approval takes the call_id of an action Axial proposed, decision (approve or reject) and an optional reason: approving runs the action as the token’s owner, who needs the right it asks for; rejecting tells Axial it was declined. chat/shared-note takes note_id and decision — send, publish, reject or discard — with optional title, body, applies_when and reason; publishing and rejecting need Manage access to Axial. See Axial shared notes.
curl -X POST https://support.sole-provider.example/api/v1/axial/chat/approval \
-H "Authorization: Bearer $TOKEN" \
-H "Content-Type: application/json" \
-d '{ "ticket_number": "528641", "call_id": "call_7Qm2", "decision": "approve" }' Steering. chat/rerun works the latest customer message again (refused once a reply has been sent for it, and while Axial is working). chat/skill takes skill_key and an optional op — add, primary or remove — and the chosen skill holds until someone changes it. chat/turn-on has Axial take the ticket’s customer messages again after a staff reply turned it off. chat/translations/approve takes a run_id and makes the stock-response translations that run produced the approved ones. It changes stock responses, so the token must also hold settings:write or templates:write. chat/skill-proposal takes proposal_id and decision (apply or discard); applying changes a skill, so the owner must be able to edit skills and the token must also hold skills:write.
Agent tasks
chat/agent-task takes an instruction and gives Axial a task on the ticket, as Give Axial a task does; chat/agent-task/deep starts the ticket skill’s in-depth investigation. The task works as the token’s owner, with a token of its own, so it is started only by a token that already holds every ability the task’s token would carry — an API call never gives a task more than the caller has. chat/agent-task/stop takes the task_id.
A token used by an agent task cannot approve an action that task proposed. Nor can Axial’s own token, or one made for one of Axial’s runs: chat/card and chat/approval refuse them with 403, so a card is answered and an action approved by a person.
Conversations with Axial
A member of staff with Work access to Axial can talk with it outside any ticket; these endpoints reach the token owner’s own conversations, and no one else’s (another person’s gives 404). Each message is answered by an agent task working as the owner, so create and message follow the agent-task rule above; reading, renaming, deleting and stopping need only the ability.
conversations— the owner’s conversations, newest first.conversations/show—conversation_id.conversations/create—message, the first message.conversations/message—conversation_idandmessage.conversations/stop—conversation_idandtask_id.conversations/update—conversation_idwithtitleto rename it, ordelete: true(not while Axial is answering).
The gap log
axial/gaps lists what Axial was asked to do and could not, with each gap’s status and owner. axial/gaps/update changes a gap’s status (open, planned, done or won’t do) with a note, its title, kind or owner, or merges it into another gap; it needs Manage access to Axial. See Axial gaps.
The MCP server’s Axial tools map onto these endpoints.