# Pptlive > LLM-friendly API/CLI for driving powerpoint. - **Type:** MCP server - **Install:** `agentstack add mcp-thomas-villani-pptlive` - **Verified:** Yes β€” security-reviewed for prompt injection and unsafe behavior - **Seller:** [thomas-villani](https://agentstack.voostack.com/s/thomas-villani) - **Installs:** 0 - **Category:** [AI & ML](https://agentstack.voostack.com/c/ai-and-ml) - **Latest version:** 0.1.0 - **License:** MIT - **Upstream author:** [thomas-villani](https://github.com/thomas-villani) - **Source:** https://github.com/thomas-villani/pptlive - **Website:** https://thomas-villani.github.io/pptlive/ ## Install ```sh agentstack add mcp-thomas-villani-pptlive ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About # pptlive [](https://pypi.org/project/pptlive/) [](https://pypi.org/project/pptlive/) [](https://github.com/thomas-villani/pptlive/blob/main/LICENSE) [](https://thomas-villani.github.io/pptlive/) Drive a running Microsoft PowerPoint instance from Python β€” `xlwings`, but for PowerPoint. Built for both human scripting and LLM agents. Windows-only. πŸ“– **[Documentation](https://thomas-villani.github.io/pptlive/)** β€” full Python API, CLI, MCP server, and cookbook. The live-app sibling of [`python-pptx`](https://python-pptx.readthedocs.io/) (which works the `.pptx` on disk) and the PowerPoint counterpart of [`wordlive`](https://github.com/thomas-villani/wordlive). Use it when the user already has the deck open and you want to edit it *live* β€” no close-the-file, let-the-agent-write, re-open dance. | Library | Target | Mechanism | | ----------- | ------------------------- | ------------------------ | | python-pptx | a `.pptx` file on disk | OOXML I/O | | **pptlive** | **a running POWERPNT.exe**| **COM automation (pywin32)** | ## Install ``` pip install pptlive # with the MCP server for Claude Desktop & other MCP agents (see "MCP server") pip install "pptlive[mcp]" # add to a project uv add pptlive ``` (Requires Python 3.11+ and `pywin32` on Windows.) ## Python ```python import pptlive as pl with pl.attach() as ppt: deck = ppt.presentations.active # Reads (structured, side-effect-free) slides = deck.slides.list() # [{index, id, layout, title, shape_count, has_notes}] outline = deck.outline() # [{slide, title, bullets:[...]}] grid = deck.slides[2].read() # every shape: anchor_id, name, id, type, geometry, text title = deck.slides[2].placeholder("title").text notes = deck.slides[1].notes.text # Polite writes β€” preserve the user's viewed slide + Selection. with deck.edit("Revise the agenda slide"): deck.anchor_by_id("ph:2:title").set_text("Agenda") deck.anchor_by_id("ph:2:body").set_text("Intro\nDemo\nQ&A") # Slide lifecycle β€” also one Ctrl-Z per edit() block. names = deck.layouts() # [{index, name}] β€” what set_layout/add accept with deck.edit("Add a results slide"): new = deck.slides.add(layout="two_content", index=4) deck.slides[7].duplicate() # copy lands at slide 8 deck.slides[9].move_to(2) deck.slides[4].set_layout("title_and_content") # Shapes & geometry β€” points throughout (pl.units for inches/cm). with deck.edit("Lay out the results slide"): shapes = deck.slides[4].shapes shapes.add_textbox("Revenue up 12%", left=pl.units.inches(1), top=72) star = shapes.add_shape("star", left=400, top=120, width=120, height=120, fill="#C00000", line="none") # fill / border on creation logo = shapes.add_picture("logo.png", left=600, top=40, # embedded, never linked alt_text="Acme logo") # a drift-proof re-id handle deck.slides[4].shapes["Picture 3"].move(top=140) # absolute, points star.set_fill(fill="#1F1F1F", line="#FFFFFF", line_width=2) # fill β‰  font color star.reorder("front") # z-order: front/back/forward/backward star.delete() # Pictures β€” alt text doubles as a re-identification handle; export one shape for vision. logo.set_alt_text("Acme logo (top-right)") # survives z-order drift chart_png = deck.slides[4].shapes["Chart 2"].export_image() # just that shape, native size # Charts β€” a chart is a shape; its data lives in an embedded Excel workbook. with deck.edit("Add a revenue chart"): chart = deck.slides[4].shapes.add_chart( "column", ["Q1", "Q2", "Q3"], {"Revenue": [10, 20, 30], "Profit": [3, 6, 9]} ).chart chart.set_type("line") # change the kind chart.recolor_text("#FFFFFF") # all chart text white (dark-theme fix) data = chart.read() # {chart_type, categories, series:[...]} # SmartArt β€” a diagram is a shape too; its content is a node tree. with deck.edit("Add a process diagram"): sa = deck.slides[3].shapes.add_smartart( "process", ["Discover", "Design", "Build", "Ship"] # flat list… ).smartart sa.set_nodes([{"text": "CEO", "children": ["VP Eng", "VP Sales"]}]) # …or a tree sa.recolor_text("#FFFFFF") # every node label white tree = sa.read() # {layout, nodes:[{text, level, children}]} # Text structure β€” paragraphs, formatting, bullets. (Per-anchor formatting; # for deck-wide styling use deck.theme / deck.master below.) with deck.edit("Polish the body copy"): body = deck.anchor_by_id("ph:4:body") body.set_text("Revenue up 12%\nChurn down 3%\nNPS +9") body.apply_list("bulleted") # bullets on every paragraph body.paragraph(2).format_paragraph(indent_level=2, alignment="left") body.paragraph(1).format_text(bold=True, size=24, color="#2E74B5") body.insert_paragraph_after("Cash runway: 30 months") # append a bullet # Tables β€” a table is a shape (Shape.has_table); cells are cell:S:N:R:C anchors. with deck.edit("Add a metrics table"): table = deck.slides[4].shapes.add_table(rows=3, columns=2).table table.cell(1, 1).set_text("Metric") table.cell(1, 2).set_text("Q3") table.add_row(["Revenue", "$4.2M"]) # appends + fills a row deck.anchor_by_id("cell:4:5:1:1").format_text(bold=True) # a Cell is an anchor grid = table.read() # {slide, shape, rows, columns, cells:[...]} # Deck-wide styling β€” theme (palette + fonts) and master (text styles + # background) restyle every inheriting slide at once. Global + anti-polite, # but still one Ctrl-Z; your view doesn't move. with deck.edit("Rebrand the deck"): deck.theme.set_color("accent1", "#C00000") # recolor the whole deck deck.theme.set_font("major", "Georgia") # major = headings, minor = body deck.master.format_text_style("body", 1, font="Georgia", size=28) deck.master.set_background("#1F1F1F") # solid fill palette = deck.theme.read() # {colors:{slot:#RRGGBB}, fonts:{major, minor}} # Find / replace β€” fuzzy traversal across shapes, table cells, and notes # (no deck-wide character stream, so this is a walk, not a range). hits = deck.find("Q3 Reuslts") # [{anchor_id, start, length, text, ...}] with deck.edit("Fix the typo everywhere"): deck.find_replace("Q3 Reuslts", "Q3 Results", all=True) # Review comments β€” slide-anchored at an (x, y) point, and threaded. review = deck.slides[2].comments # per-slide CommentCollection with deck.edit("Leave review notes"): c = review.add("Tighten this headline", left=100, top=80) c.reply("Agreed β€” will do") roll = deck.comments() # deck-wide roll-up {total, slides:[...]} # Whole-deck snapshot β€” one low-res PNG per slide so a vision model can SEE # the whole deck cheaply (max_dim caps each slide's long edge). A read β€” polite. snaps = deck.snapshot(max_dim=1000) # [Snapshot(slide, image, path), ...] # Output tier β€” explicit, never implicit (pptlive never auto-saves). deck.save() # persist to the existing file deck.save_as("v2.pptx", overwrite=True) # write + rebind the working file deck.export_pdf("deck.pdf") # a read: no rebind, dirty flag kept # Media + narrated-video export β€” build a deck, narrate it, export an MP4. with deck.edit("Narrate the deck"): deck.slides[1].add_audio("intro.mp3") # embed; autoplay + pace the slide deck.slides[2].add_video("demo.mp4") # a video clip (stays visible) result = deck.export_video("deck.mp4", resolution=1080) # async CreateVideo; blocks to done assert result.ok and result.status == "done" # result.path is the MP4 # Render β€” let a vision model *see* the slide it just built (export β†’ read β†’ iterate). png = deck.slides[4].export_image(width=1280) # temp PNG (or pass a path); polite # ...hand `png` to your image tool, look, then revise. # Read what the user is looking at, and (opt-in) target it with the here: anchor. sel = deck.selection() # {type, slide, anchor_id, shapes, ...} if sel.anchor_id: with deck.edit("Bold the selected text"): deck.anchor_by_id("here:").format_text(bold=True) # Live slide show β€” drive the presentation like a clicker (deliberately moves the screen). deck.show.start() # run from the top deck.show.goto(5); deck.show.next(); deck.show.black() # jump, advance, blank deck.show.state() # {running, state, current_slide, ...} deck.show.end() ``` ## Anchors Addressing is **hierarchical** (slide β†’ shape β†’ text), not a global character stream β€” there is no deck-wide `range:`. Anchor ids are colon-separated, slide-index first: | anchor_id | resolves to | | -------------- | ----------- | | `shape:S:N` | Nth shape (1-based z-order) on slide S β€” the canonical handle | | `shapeid:S:ID` | shape with stable `Shape.Id` ID on slide S β€” the **delete-proof** handle (survives a delete/restack that shifts `shape:S:N`) | | `ph:S:KIND` | placeholder of semantic KIND (`title`/`ctrtitle`/`subtitle`/`body`/`footer`/`date`/`slidenum`) β€” the LLM-preferred form | | `para:S:N:P` | paragraph P (1-based) of shape N on slide S | | `cell:S:N:R:C` | cell (row R, col C) of the table in shape N on slide S β€” a `Cell` *is* an anchor, so it takes every text/format verb | | `notes:S` | speaker-notes body of slide S | | `comments:S` | the review comments on slide S β€” a read selector (a container, addressed for reply/delete by `(slide, 1-based index)`) | | `here:` | whatever the user has selected right now β€” the shape, or the paragraph holding the text caret (the opt-in way to act on the live selection) | z-order **drifts** when shapes are added or removed (or restacked via `Shape.reorder`), so `shape:S:N` is resolved live and never cached; every shape listing also emits `name` (`Shape.Name`) and `id` (`Shape.Id`, stable across reorder). The drift-proof forms are `ph:S:KIND`, `.Name`, and `shapeid:S:ID`. `para:S:N:P` and `cell:S:N:R:C` also resolve live (the paragraph/row count shifts as text or rows are inserted/deleted). ## CLI JSON in, JSON out, deterministic exit codes β€” designed to drop straight into an LLM tool-use loop. Global flags (`--json`/`--text`, `--doc NAME`) go *before* the subcommand. ``` pptlive status # open decks, active one, slide in view pptlive slides # [{index, id, layout, title, shape_count, has_notes}] pptlive outline # title + body bullets per slide pptlive slide read 2 # every shape on slide 2 pptlive shapes --slide 2 # shapes on slide 2 (anchor_id, name, id, type, geometry) pptlive read anchor --anchor-id ph:2:title # read any text anchor (ph:/shape:/notes:) pptlive read notes --slide 1 # sugar for --anchor-id notes:1 pptlive write --anchor-id ph:2:body --text "Intro\nDemo\nQ&A" pptlive replace --anchor-id shape:3:1 --text "New text" pptlive find --text "Q3 Reuslts" [--in slide:2] # fuzzy locate across shapes/cells/notes pptlive replace --find "Q3 Reuslts" --text "Q3 Results" --all # fuzzy replace (or --occurrence N) pptlive slide layouts # the layout names add/set-layout accept pptlive slide add --layout two_content [--index 4] pptlive slide duplicate --slide 7 pptlive slide move --slide 9 --to 2 pptlive slide set-layout --slide 4 --layout title_and_content pptlive slide delete --slide 5 pptlive slide export --slide 2 --out slide2.png [--width 1280] [--format png] # render to image pptlive slide geometry 2 # slide size + shape boxes + overlaps + off-slide (no render) pptlive shape add --slide 4 --kind textbox --text "Revenue up 12%" --left 72 --top 72 pptlive shape add --slide 4 --kind shape --shape-type star --left 400 --top 120 --width 120 --height 120 pptlive shape add --slide 4 --kind picture --path logo.png --left 600 --top 40 --alt-text "Acme logo" pptlive shape add --slide 4 --kind table --rows 3 --cols 2 --left 72 --top 120 pptlive shape move --anchor-id shape:4:3 --left 100 --top 140 pptlive shape resize --anchor-id shape:4:3 --width 300 --height 200 pptlive shape delete --anchor-id shape:4:3 pptlive shape set-alt --anchor-id shape:4:3 --alt-text "Acme logo" # drift-proof re-id handle pptlive shape fill --anchor-id shape:4:3 --fill "#C00000" --line none --line-width 2 # fill/border pptlive shape order --anchor-id shape:4:3 --to front # z-order: front/back/forward/backward pptlive shape export --anchor-id shape:4:3 --out logo.png # render one shape (native size) pptlive shape animate --anchor-id shape:4:3 --effect fly_in [--trigger after_previous] [--exit] pptlive slide animations 4 # a slide's shape animations in play order pptlive shape add --slide 4 --kind chart --chart-type column \ --categories "Q1,Q2,Q3" --series '{"Revenue":[10,20,30]}' pptlive chart read --slide 4 --shape 5 # {chart_type, categories, series} pptlive chart set-type --slide 4 --shape 5 --chart-type line pptlive chart set-data --slide 4 --shape 5 --categories "A,B" --series '{"S":[1,2]}' pptlive chart recolor-text --slide 4 --shape 5 --color "#FFFFFF" # all chart text (dark-theme fix) pptlive shape add --slide 3 --kind smartart --smartart-kind process \ --nodes '["Discover","Design","Build","Ship"]' # flat list or {text,children} tree pptlive smartart read --slide 3 --shape 2 # {layout, nodes:[{text, level, children}]} pptlive smartart set-nodes --slide 3 --shape 2 --nodes '[{"text":"CEO","children":["Eng","Sales"]}]' pptlive smartart recolor-text --slide 3 --shape 2 --color "#FFFFFF" # every node label pptlive theme read # deck palette (12 slots) + heading/body fonts pptlive theme set-color --slot accent1 --color "#C00000" # recolors the whole deck pptlive theme set-font --which major --name "Georgia" # major = headings, minor = body pptlive master read # master text styles (title/body/default) + background pptlive master format-text-style --style body --level 1 --font "Georgia" --size 28 pptlive master set-background --color "#1F1F1F" # deck-wide; solid fill pptlive paragraphs --anchor-id ph:4:body # [{anchor_id (para:S:N:P), text, indent_level, bullet}] pptlive insert --anchor-id para:4:2:3 --text "New bullet" [--before|--after] pptlive format-paragraph --anchor-id para:4:2:1 --alignment center --indent-level 2 pptlive format-text --anchor-id ph:4:title --bold --size 40 --color "#2E74B5" pptlive list apply --anchor-id ph:4:body --type bulleted [--char "β€’"] pptlive list remove --anchor-id ph:4:body pptlive table read --slide 4 --shape 5 # grid of cells, each with its cell:S:N:R:C anchor pptlive table add-row --slide 4 --shape 5 --values '["Revenue", "$4.2M"]' pptlive table delete-row --slide 4 --shape 5 --row 2 pptlive write --anchor-id cell:4:5:1:1 --text "Metric" # a cell takes write/format-text/... pptli … ## Source & license This open-source MCP server is cataloged on AgentStack and links to its original source β€” we do not rehost the code. - **Author:** [thomas-villani](https://github.com/thomas-villani) - **Source:** [thomas-villani/pptlive](https://github.com/thomas-villani/pptlive) - **License:** MIT - **Homepage:** https://thomas-villani.github.io/pptlive/ Install and usage instructions live in the source repository linked above. ## Pricing - **Free** β€” Free ## Security capabilities Automated source analysis of v0.1.0 β€” what this tool can access: - **Network access:** no - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** no - **Dynamic code execution:** no *"Yes" means the capability is present in the source β€” more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** β€” security scan: passed β€” Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/mcp-thomas-villani-pptlive - Seller: https://agentstack.voostack.com/s/thomas-villani - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack β€” the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.