Incident postmortem drafter
By Toolspoke
Drafts a Confluence postmortem from each resolved incident.io incident, with its timeline, impact and follow-ups, for a person to finish.
Needs a Toolspoke account. You choose which project it installs into.
- incident.io + Confluence
- Required integrations
- Runs on a schedule
- Every day at 10am by default
- Say what you did
- What its last step does
- Some actions can delete
- 2 of 29 actions in its tools
On this page
- The short version
- What it will never do
- What Incident postmortem drafter does
- How it works
- What a drafted postmortem looks like
- Who uses Incident postmortem drafter
- What it connects to
- When it runs
- How to set it up
- Make Incident postmortem drafter fit your team
- Compared with other options
- Questions
- Related agents
The short version
- What it is
- A ready-made Toolspoke agent that drafts a postmortem in Confluence for each incident resolved in incident.io.
- What it reads
- Each incident's details, every status update and its follow-ups.
- What it writes
- One page under the parent you chose, with the timeline, impact, causes as stated and follow-ups.
- What you get
- A postmortem that is mostly assembled the morning after, marked [Draft] for the incident lead to finish.
- How often
- Every day at 10 AM once you switch it on, or for one incident whenever you ask.
- What it never does
- Edit an incident, create a follow-up, or state a cause nobody wrote.
- What you need
- A Toolspoke account, an incident.io API key, and a Confluence space with a parent page for postmortems.
What Incident postmortem drafter will never do
Incident postmortem drafter is told never to change an incident, never to state a cause nobody wrote, and never to publish outside the parent page you chose. It drafts, and a person finishes.
Incident postmortem drafter will never
- Edit an incident, post an update on one, change its severity or status, or create, edit or close a follow-up.
- Declare an incident or page anyone.
- Create a page anywhere but under the chosen parent, update or move an existing page, or remove "[Draft]" from a title.
- State a cause, a root cause or a lesson that no responder wrote.
- Name a person as the cause of anything.
- Copy customer names, email addresses, credentials or personal data from updates into the page.
Why it will not touch the incident
An incident update goes to everyone in the incident channel, and a follow-up lands on someone's list. Those are the responders' to make. With a View data key, the agent cannot make them.
Why it will not supply a cause
A plausible cause written by an agent is easy to accept and hard to question. The draft quotes what responders stated and leaves the analysis to the team.
Why it will not name a person
A postmortem that blames a person stops people writing honest updates. Roles and follow-up owners are listed as incident.io records them, and nobody is named as a cause.
These are written into Incident postmortem drafter's instructions. They guide the agent, but they are not a permission. What it is allowed to touch is set by the access you grant.
Incident postmortem drafter assembles the record. What went wrong, what to change and when to publish stays with your team.
What Incident postmortem drafter does
Incident postmortem drafter is an AI agent that finds the incidents resolved in incident.io in the last day and drafts a postmortem page for each one in Confluence. It assembles the record. Your team writes the conclusions.
A postmortem is easiest to write the day after and hardest to start. Someone has to scroll back through every update, copy the timestamps, list the follow-ups and get the impact right before they can think about what went wrong.
That assembly is the slow part, and it is all in incident.io already. The thinking is the part only the team can do.
Incident postmortem drafter does the assembly. It reads the incident, every update and every follow-up, writes the timeline in UTC, quotes what responders said about the cause, and leaves clear headings for the sections a person has to write.
In practice, the morning after an incident you have:
- A Confluence page under your postmortems parent, titled [Draft]
- What people experienced and for how long
- A timeline with every update and status change, in UTC
- The causes responders stated, each with the update it came from, and the follow-ups as recorded
How it works
Incident postmortem drafter works in four steps each morning: it finds the resolved incidents, checks for an existing draft, reads each incident and writes the page. All of it comes from the agent's written instructions, which you can read and edit after you install it.
Step 1: Find resolved incidents
Those resolved in the last 24 hours, at the severities you chose.
Step 2: Check for a draft
Search the space for the incident's reference first.
Step 3: Read the record
The incident, every update and every follow-up.
Step 4: Write the draft
One page under your parent, marked [Draft].
It copies causes, it does not find them
Under Causes as stated, the draft quotes what responders wrote, each line with the time of its update. If nobody stated a cause, it says so. The analysis is a heading for the team to fill.
It never doubles up
Before writing, it searches the space for the incident's reference with search_pages. If a page exists, it links it and drafts nothing.
It marks the page unfinished
The title starts with [Draft] and the page opens with a note saying it was drafted from incident.io records and not reviewed. It is a normal page, not a Confluence draft, because a Confluence draft is seen only by the account that saved it.
It does not touch the incident
It never edits an incident, posts an update or adds a follow-up. With a key that has only View data, it cannot. The limits are listed under what it will never do.
What a drafted postmortem looks like
A run of Incident postmortem drafter writes one Confluence page per resolved incident. The page leads with what the team still has to write, then the assembled record.
[Draft] Postmortem: INC-214 Checkout payments timing out
Drafted by an agent from incident.io records on 28 October. Not reviewed. Causes are as responders stated them. The incident lead finishes this page.
For the team to write (4)
- Contributing factors: what made this possible, beyond the trigger.
- Why detection took 14 minutes: from the first error to the incident being declared.
- Why repair took 38 minutes: from declaration to resolution.
- Lessons: what the team would do differently.
The record
- What people experienced: for 52 minutes, about one in four card payments at checkout timed out. Severity 1.
- Timeline: 17 entries, from 14:02 UTC (first update, investigating) to 14:54 UTC (resolved).
- Causes as stated: "Connection pool on the payments service exhausted after the 13:40 deploy raised worker count" (update at 14:31 UTC).
Follow-ups (3)
Cap pool size per worker (Payments team, open). Alert on pool saturation (Platform, open). Add load test for checkout (Payments team, open).
- For the team to write
Empty headings for contributing factors, detection and repair time, and lessons, each with one line saying what belongs there. This is the part that needs judgement.
- The record
Impact, duration, severity, the timeline in UTC and the causes responders stated, all taken from incident.io.
- Follow-ups
Each follow-up already recorded in incident.io, with its owner and status, exactly as they are there.
When nothing was resolved
If no incident was resolved in the last day at the severities you chose, it creates nothing and says so in the run result.
If Confluence cannot be written, the draft goes in the run result so the work is not lost. If an incident has no updates, the timeline is drawn from its timestamps alone, and the page says so.
Who uses Incident postmortem drafter
Incident postmortem drafter suits an engineering team that runs incidents in incident.io and keeps postmortems in Confluence. The examples use made-up names.
An incident lead the morning after
Sofia Lindqvist led last night's checkout incident. She opens the draft, fills in the four headings and publishes by lunch.
An engineering manager tracking reviews
Arjun Mehta checks the postmortems parent each week. Every resolved Sev 1 and Sev 2 has a page, so none are forgotten.
A responder checking the record
Chen Wei was on call. He reads the timeline to check his updates are in the right order before the review meeting.
A reliability lead running the review
Fatima Zahra starts the review from the stated causes and follow-ups, rather than a blank page.
What Incident postmortem drafter connects to
Incident postmortem drafter connects to two tools: it reads incident.io and writes pages in Confluence. You connect each one yourself, and Toolspoke stores the connection encrypted, so the agent never sees the credential.
incident.io
RequiredRead and run incidents, status updates, follow-ups, alerts, on-call schedules and escalations in incident.io.
Authentication: API key
Confluence
RequiredSearch the wiki, read pages, publish and revise documentation.
Authentication: Atlassian API token
| Tool | What Incident postmortem drafter does | What it cannot do |
|---|---|---|
| incident.io | Lists incidents and severities, and reads each incident, its updates and its follow-ups (list_incidents, list_severities, get_incident, list_incident_updates, list_follow_ups). | Edit an incident, post an update, add a follow-up, declare an incident or page anyone, if the API key has only View data. Each of those needs its own permission on the key. |
| Confluence | Finds the space, checks for an existing draft and creates one page per incident (list_spaces, search_pages, create_page). | Delete a page: the connector has no delete action. It can update pages and comment, and the agent is told only to create pages under the parent you chose. |
That the agent cannot change an incident is a lock, if you create the incident.io key with only View data. That it only creates pages under one parent in Confluence is an instruction to the agent, not a lock. If you want a lock, a project can switch individual actions off. The access control guide explains how.
The Confluence token carries the permissions of the account that made it. Use an account that can create pages in the postmortems space, and the draft is written as that account.
When Incident postmortem drafter runs
Incident postmortem drafter runs every day at 10 AM, but only after you switch its schedule on. It arrives switched off, so nothing runs and nothing is charged until you decide. You can also run it on demand for one incident.
- Each run finds incidents resolved in the last 24 hours.
- It drafts one page per incident, under the parent you chose.
- It skips an incident that already has a page.
- If nothing was resolved, it creates nothing.
Each run uses model credit from your workspace balance, so it makes one run a day, and a run on a day with no resolved incidents is short. The Toolspoke tool calls themselves are free.
Your copy is an ordinary agent in your project. You can pause its schedule there, run it once on demand, change the time and edit its instructions. The agent's page lists recent runs and their cost.
How to set up Incident postmortem drafter
To start drafting postmortems, add Incident postmortem drafter to a Toolspoke project, connect incident.io and Confluence, choose the parent page and switch on the schedule. It takes a few minutes, and nothing runs until the last step.
Step 1: Create a Toolspoke account or sign in
Choose Use this agent. If you are new, you will be asked to sign up first, and you come straight back to this agent afterwards.
Step 2: Add it to a project
Pick the project it should live in. You get your own editable copy, and you can see the two tools it brings before anything connects.
Step 3: Connect incident.io with View data
In incident.io, add an API key and tick View data only. That is all the agent needs, and it means the key cannot change an incident.
Step 4: Connect Confluence
Enter your site name and an API token from an account that can create pages in your postmortems space.
Step 5: Choose the parent page and severities
Name the space and the parent page drafts go under, and the severities that earn a postmortem.
Step 6: Switch on the schedule
The schedule arrives switched off. Run it once on a recent incident, read the draft, then switch it on.
Make Incident postmortem drafter fit your team
Incident postmortem drafter's instructions are plain English and you can edit them after you install it. Installing copies the agent, so your edits stay with your copy. Open the agent in Toolspoke and tell the builder what to change, and it rewrites the instructions.
Paste any sentence below into the builder chat, or say the same thing in your own words.
Use these headings, in this order: Summary, Impact, Timeline, Contributing factors, Action items.
Draft a postmortem only for Sev 1 and Sev 2 incidents.
Create the page with status draft instead of current.
A Confluence draft is seen only by the account that saved it, so share it before anyone else can read it.
End the page with a heading Review meeting, with the date left blank for the incident lead.
None of these edits needs more access or touches the Never rules. They change what Incident postmortem drafter is told, not what the connections allow.
How it compares to other options
Incident postmortem drafter beats the alternatives when you write postmortems in Confluence and want the record assembled the morning after, and loses to them when you want the whole review written for you.
| Compared on | Incident postmortem drafter | incident.io's own postmortems | A Confluence template | Writing it from scratch |
|---|---|---|---|---|
| Lands in Confluence | Yes | Its own page | Yes | Yes |
| Timeline filled in | Yes, from every update | Yes | No | By hand |
| Follow-ups listed | Yes | Yes | No | By hand |
| Started without anyone asking | Every morning | If set up | No | No |
| Writes the analysis | No, leaves it to you | Varies by plan | No | Yes |
| Cost | A run a day | Your incident.io plan | Free | Hours per incident |
incident.io's own postmortem feature is the better choice if your team is happy to keep postmortems in incident.io. Incident postmortem drafter is for the team whose reviews live in Confluence, next to the rest of its documentation.
Writing it from scratch gives you the most control, and the draft does not take that away. It does the copying, so the person writing can start at the analysis.
Frequently asked questions
Can Incident postmortem drafter change an incident?
Not if you create the incident.io key with only View data. Every write and every page needs its own permission on the key. The agent is also told never to use them.
Does it work out the root cause?
No. It quotes what responders wrote about the cause, each with the time of its update. If nobody stated one, the draft says so. The analysis is for your team.
Why is the page not a Confluence draft?
Because a Confluence draft is seen only by the account that saved it, and your team needs to find the page. The [Draft] title and the note at the top mark it unfinished. You can ask for real Confluence drafts by editing the instructions.
Will it create a second page for the same incident?
No. It searches the space for the incident's reference first, and if a page exists it links it and drafts nothing.
Can I run it for one incident?
Yes. Run it on demand and give it the incident, and it drafts that one, whenever it was resolved.
Does it work with Notion or Google Docs?
Not as shipped, because it writes to Confluence. You can ask the agent builder for a version that drafts elsewhere. It needs its own instructions.
Is Incident postmortem drafter free?
It has no price of its own, and every Toolspoke plan includes agents. Tool calls cost nothing, but each run draws on your usage balance. Pricing has the details.
Start every postmortem half written
Connect incident.io and Confluence, choose the parent page and switch on the schedule. The next resolved incident gets a draft the morning after.