Loading page
About this tool
Turn long email conversations into a concise local summary with decisions, action items, participants, and open questions.
The Email Thread Summarizer splits a pasted email chain into its individual messages, then extracts decisions, action items with owners and deadlines, unanswered questions and a ranked set of highlights using keyword patterns rather than a language model. It is for anyone who has been added to a 30-message thread and needs to know what was agreed and what is now theirs to do. Because every rule is a local regex, the same thread always produces the same summary, and the text never leaves the browser.
Open Email Thread Summarizer on AltFTool — it loads instantly in your browser.
Paste your chain (up to 100,000 characters) into the "Paste email thread" box — From, Date and Subject headers improve results.
Pick a Summary length of Brief, Standard or Detailed, which keep the top 3, 6 or 10 scored sentences alongside decisions, action items and questions.
Click Copy for the analysis as Markdown, or download it as email-analysis.md or email-analysis.json.
It splits on From/To/Subject/Date headers, drops quoted lines starting with > and strips sign-offs like 'Best regards', so replies are not counted twice.
Each action item is matched to a participant named in the text, or to the sender for 'I will' and to Team for 'we will', with any date phrase captured separately.
No model call and no randomness — the same thread yields the same decisions, questions and highlights every time, and nothing is uploaded.
Each sentence gets a score and the top ones are kept: +3 if it matches an action pattern, +3 for a decision pattern, +2 for a question, +2 for being the opening sentence of a message, +1 for words like because, however, deadline, budget, risk or priority, plus a small bonus for appearing later in the thread. Brief keeps the top 3, Standard 6 and Detailed 10.
Up to 12 action items, 10 decisions and 10 questions per thread, after near-duplicate sentences are collapsed. Action items are sentences containing cues such as please, need to, should, must, can you, could you, follow-up or due.
No. Parsing, scoring and summarising all run in your browser with regular expressions and a scoring loop, so the thread is never transmitted. Copy and the Markdown and JSON downloads are generated locally too.
Treat it as a prompt, not a verdict. A question is marked 'Possibly answered' only when a later message contains words like yes, no, confirmed, agreed, approved or will — the tool cannot tell whether that reply addressed the specific question, so check the original thread before acting on anything consequential.