Your team's work log, written by their commits.
Devlog reads the commits your team already pushed, matches them to the tasks they were assigned, and writes each person's daily log. Managers get a timesheet nobody had to type.
- Public + private repos
- Bring your own AI provider
- Local models via Ollama
- Every entry is a draft until you submit it
- $
Today · nightly run · 22:00
draftCommits in from · Logs out to
- GitHub
- GitLab
- Bitbucket
- Azure DevOps
- Jira
- Linear
- Notion
- Sentry
- Google Calendar
- Outlook Calendar
- Slack
- Discord
- Microsoft Teams
- OpenAI
- Anthropic
- Google Gemini
- Groq
- Mistral
- Ollama · local
- Any OpenAI-compatible endpoint
- Google Sheets
- SQL Server
- Supabase
- Custom webhook
0
AI providers
cloud, local, or your own endpoint
0+
sources and sinks
commits, issues, calendars, sheets
0%
yours to review
every entry is a draft until you submit it
0
standup notes typed
the log is already written
What it does
Everything the log needs, without the typing.
Stop context-switching between your editor, GitHub, and the team sheet. Devlog watches the work and writes it down.
Commits from everywhere you work
Public and private repos across GitHub and GitLab, plus the work that never produced a commit — reviews you left, PRs you opened, issues you closed.
The AI provider you choose
OpenAI, Anthropic, Gemini, Groq, Mistral, a local model through Ollama, or any OpenAI-compatible endpoint. Your key, your model, your prompt.
provider = "OpenAI"
key = "yours"
prompt = "yours too"
Drafts until you submit
Summaries you generate are shown before they're saved; nightly runs land as drafts for you to review.
Hours that say where they came from
Measured, typed, or estimated — and the entry always says which.
Back-fill a missed week
Pick a range and every day with activity and no entry gets filled in. Logged days are left alone.
Runs while you sleep
Set a time and a timezone. 22:00 means 22:00 where you are, not where the server is. A failed night can be retried the same day.
Asia/Kolkata
● next run in 3h 12m
retry same day if it fails
Local, if you want it
Self-host next to Ollama and commit messages never leave your own infrastructure.
0 bytes leave your servers
How it works
Set up once. Then it runs nightly.
Four steps, one evening. After that the log is waiting for you each morning.
Step
01
Set a time and a timezone once. 22:00 means 22:00 where you are, and every morning the log is already there.
- Step 1
Connect GitHub
OAuth in about a minute, or paste a token if your org restricts OAuth apps. Minimal scopes — commit history and your email.
Continue with GitHubscopes: repo, user:email
- Step 2
Assign the work
Create a project, give it a key, point it at your repos. Assign tasks to people. Solo? Skip this entirely — it's optional.
Todo
DEV-42
Branch filter on commit fetch
DEV-43
Weekly digest opt-in
Doing
DEV-41
Rate-limit the ingest route
Done
- Step 3
Pick your model
A cloud provider with your own key, or a local one through Ollama. Your prompt, your tone, your data.
OpenAI · gpt-4.1Anthropic · claude-sonnet-5Ollama · llama3 (local) - Step 4
Review and approve
Read what the model wrote, fix anything that misreads the day, and approve. Only then does it become an entry.
Shipped the branch-aware commit fetch and covered the timezone rollover with tests.
6.5h · measured · DEV-42Approve
For engineering teams
The log writes itself against the task you assigned.
Put a task key in a commit message — DEV-42 — and that day's log is linked to that task, exactly, with no AI guesswork involved. Commits that mention nothing still get summarized; the entry records which links were certain and which were inferred.
Managers stop chasing standup notes. Engineers stop writing them.
commit · a41f2c0
task · DEV-42
log · today
Invite your engineers
Seats sized to your headcount. Owners, admins, managers, members — each sees exactly what their role warrants.
Assign the work
Projects and tasks with a board, comments, and an audit trail. Each project names the repositories its work happens in.
Read the report, not the repo
Every member's logged days in one table, with hours and provenance, exportable as CSV for whatever finance uses.
Separated at the database
One company's data is unreachable from another's by row-level security, not by a filter in application code someone could forget.
Pricing
Priced by how many people you have.
Solo is free, forever. Everything above it is per seat, and includes a shared pool of AI credits each month.
Credits pay for Devlog's pipeline: about 1.5 per summarized commit, and only for days that produce an entry. Model calls are made with your own provider key and billed by that provider.
FAQ
Questions people actually ask.
It reads commit messages, file counts, and line counts — not file contents. Those are what get sent to whichever model you configured. Self-host Devlog next to a local Ollama model and nothing leaves your own infrastructure at all.
Spend tonight shipping. Not writing standup notes.
Connect GitHub, pick a model, and tomorrow morning the log is already written — waiting for you to read it rather than write it.