What is Coda: the doc that thinks it's a database

TL;DR. Coda is a doc-and-database hybrid — documents with tables that act like databases, formulas everywhere, buttons that trigger automations, and Packs that pull data from other tools. It launched in public beta in 2017, was acquired by Grammarly in December 2024, and now lives inside the company that rebranded itself Superhuman in October 2025. This post walks through what Coda actually is, the formula-and-Packs mental model, where it’s genuinely strong, where the abstraction starts costing, and whether your team is in the Coda is exactly right camp or the we should probably just have a wiki and a database camp.

Coda is the tool that says we are the document and means something specific by it. Not a wiki, not a spreadsheet, not quite Notion — a doc with the powers of an app. The pitch is genuinely interesting, the execution is genuinely careful, and the market position in 2026 is genuinely strange, because Coda is now part of Superhuman, which used to be Grammarly, which acquired Coda about eighteen months ago. The post that explains what is Coda in 2026 has to explain the corporate context too, and most don’t.

We’ll cover what Coda actually is, how the blocks-tables- formulas-Packs mental model works, the Grammarly-to- Superhuman corporate story, Coda AI, where Coda is strong, where the abstraction breaks, and the test you can run to decide whether the one tool to replace them all pitch matches your team or not.

What Coda actually is

Coda is a doc-and-database editor. The full sentence is Coda is a hosted, multi-user document editor that combines text, structured tables, formulas, buttons, and third-party integrations (Packs) into a single building-block system, designed around the editorial idea that documents and apps should be the same thing. The short sentence does most of the work; the long sentence explains why people either love it or eventually outgrow it.

Coda Project, Inc. was founded in 2014 by Shishir Mehrotra and Alex DeNeui (they met at MIT). The product entered public beta in October 2017, exited beta in February 2019, hit unicorn status in 2021, and was acquired by Grammarly in an all-stock deal in December 2024. In October 2025 Grammarly rebranded itself as Superhuman, with Coda as a core piece alongside Grammarly’s writing tools, the Superhuman email client, and a new AI assistant. The product’s product page still says Coda; the company that ships it now answers to Superhuman.

For the broader workspace category Coda sits inside, see our Notion explainer, the Confluence explainer, and the corporation wiki explainer. Coda is a shape of the category; the category does not collapse to Coda.

How Coda works

The mental model is bigger than Notion’s, smaller than Confluence’s, and lives mostly at one level: blocks live in docs; docs contain tables; tables can be formula-driven; buttons act on tables; Packs talk to other systems.

Blocks. Like Notion, every piece of a Coda page is a block — paragraphs, headings, callouts, images, tables, embedded widgets. Blocks compose. Unlike Notion, the editor leans toward the doc end of the spectrum first; the database surface lives inside the doc rather than alongside it.

Docs. The unit of organisation. A Coda doc is a single container with a tree of sub-pages, its own sharing settings, its own published-URL, and (this is the interesting part) its own runtime environment. Tables live in docs. Formulas live in docs. Buttons live in docs. The doc is the app surface; the page hierarchy is mostly a navigation aid.

Tables. The unit Coda is genuinely architecturally distinct on. A Coda table looks like a spreadsheet, behaves like a database, and is embedded in a doc. Columns are typed (text, number, date, person, image, lookup, formula). Rows are the records. The same table can be displayed as a table view, a card view, a calendar view, a Gantt chart, or a chart — the views are projections of the underlying table, not separate objects.

Formulas. Coda Formula Language is the system’s spine. A formula can compute a cell, populate a column, control whether a button is enabled, pick which view to show, or generate the contents of a synced sub-page. Formulas in Coda are not optional — they are the abstraction that turns a doc into an app. Bandit, our CTO, called Coda formulas “a quantum spreadsheet that happens to be a doc” and we have not been able to find a better description.

Buttons and automations. A Coda button runs an action — write a row, push to a Pack, send a Slack message, generate a sub-page from a template. The doc starts behaving like a small app once you have three or four buttons. Coda automations extend the model to triggered actions (on a schedule, on a webhook, on a row change).

Packs. The integration layer. Coda Packs (450+ of them as of early 2026) are pre-built integrations that pull data from or push data to external services — Salesforce, Slack, GitHub, Jira, Notion, Stripe, Google Calendar, the list goes on. A Pack can populate a table, return a value to a formula, or trigger an action from a button. Packs are the part of Coda that scales the doc-as-app pitch to real apps.

The Grammarly / Superhuman story

This is the part most explainers undercover, and as of mid-2026 it matters.

Coda was acquired by Grammarly in December 2024 for an all-stock deal of undisclosed size. In October 2025 Grammarly the company rebranded itself as Superhuman — the email-client acquisition Grammarly had folded in earlier became the brand of the holding company. Coda is now one of three products inside Superhuman: the original Grammarly writing assistant, the Superhuman email client, and Coda. There is also a Superhuman AI assistant layered across all three.

For Coda users, the practical effects so far are: the product roadmap is paced with the Grammarly/Superhuman cadence (less frequent than indie Coda’s, more deliberate), the AI surface across Coda + Grammarly + Superhuman is being unified, and the small-team free tier has stayed where it was. The strategic effects are harder to read. Companies bought into larger holding companies sometimes thrive (Notion stayed independent; Coda did not). The relevant question for a buying team is whether the team is comfortable adopting a tool whose long-term roadmap is one of three priorities for its new parent.

Coda AI

Coda AI is the assistant integrated across the editor. Summarise this table, write a status update from these rows, suggest a formula for this column, draft a meeting agenda from this doc — the standard 2026 doc-AI surface. The formula-suggestion piece is the most Coda-specific: formulas are hard, and an AI that can read your table schema and propose a working formula is a meaningfully better tool than the formula editor by itself.

Coda AI is gated to paid plans — Pro, Team, and Enterprise — with usage credits per Doc Maker. The free tier ships without Coda AI. The pricing detail moves enough that we are not going to quote it in a post that ships once; Coda’s pricing page is the only authoritative source. (Note also that under the Superhuman umbrella, the AI surface for Coda is being unified with Grammarly’s and Superhuman’s. The pricing implications of that are still developing.)

What Coda does well

Honest section. Things Coda is genuinely good at, with no caveats.

The doc-as-app metaphor. Coda’s central editorial idea — the doc is the app — is one of the best-executed abstractions in the productivity category. A working Coda doc is something between a Google Doc and a small custom- built tool, and the transition from one to the other happens gradually inside a single artefact.
Tables as a first-class primitive. Coda tables are genuinely well-designed — typed columns, reactive views, in-line filters, lookups across tables. For teams that have been doing real work in spreadsheets and want to upgrade, Coda is one of the best landing zones.
The formula language. Coda Formula Language is expressive, well-documented, and consistent in a way most no-code formula systems aren’t. The learning curve is real; the language pays back.
Packs. 450+ integrations is a serious investment, and the Pack catalogue is one of the practical reasons Coda competes with workflow tools that aren’t even document editors.
Sharing and publishing. A Coda doc can be shared internally, published externally, or embedded in another product. The published-doc surface is clean and is a real alternative to building a small marketing page or status dashboard.
The “What I Run On Coda” community. Coda’s community of Doc Makers (their term for power users) is unusually active. Template galleries, community-built Packs, and a Slack of professionals who use Coda the way other professions use Excel. This is the part the SERP routinely undersells.

Where Coda falls short

Equally honest section. Things Coda does badly, with the same lack of caveats.

Performance at scale. Coda’s doc-as-app abstraction is expensive. A doc with several large tables, complex formulas, and a few Packs loads visibly slower than the pure-document equivalent. The pattern shows up at the same scale Notion struggles at, for similar reasons.
The mental model is heavy. Doc and table and formula and button and Pack and view and sync are seven concepts a new user has to internalise before Coda becomes more useful than Google Docs. New joiners on a Coda-heavy team spend a real week learning the surface.
Search. Coda search has improved meaningfully but is still optimised for inside-this-doc search; cross-doc search at scale is the same where do new docs live problem we wrote about in the Notion post.
No native desktop app. Coda is a browser-based product (with mobile apps for iOS and Android). For users who expect a desktop app the way Notion and Obsidian ship one, this is a recurring point of friction.
Coda AI is paywalled. The AI surface — the part Coda is marketing hardest — lives on paid tiers with credit limits. A team that signed up for Coda because of the AI demo will find the meter quickly.
The Superhuman roadmap question. This isn’t a feature problem; it’s a strategic-uncertainty problem. The product is now one of three under a parent that rebranded twelve months ago. The long-term roadmap is whatever Superhuman decides it is.
Export and migration. Coda exports to Word, PDF, and a CSV-per-table format. None of the export shapes round-trip cleanly — the doc-as-app structure is the kind of thing that doesn’t survive a re-import into a different tool. The migration is the moat.

Coda’s formula language is the feature and the failure

This is the opinion section, and we’re going to keep it brief because Coda users have already had the argument with themselves a dozen times.

Coda’s formula language is the feature and the failure. The same abstraction that lets a single Doc Maker build a CRM, an editorial calendar, and an OKR tracker in one weekend is the abstraction the rest of the team can’t easily edit. Coda docs concentrate authority in whoever wrote the formulas. The team without a Doc Maker has flat docs; the team with one has a custom app exactly one person can maintain. Notion’s flexibility is the feature and the failure in one shape; Coda’s formula language is the feature and the failure in the other.

The fix isn’t to make Coda less powerful. The fix is to use Coda where the formula-as-power makes sense (small teams where one Doc Maker is the right shape, single-purpose internal tools) and to use a wiki where the team — not the Doc Maker — needs to write the docs. A wiki has spaces, labels, an activity feed, and notifications for pages you follow. Coda has formulas, Packs, and Doc Makers. Sub-second loads. Keyboard-first. For the team-wiki side of the question, see our knowledge management software guide and the corporation wiki explainer.

When to stay on Coda and when to leave

The decision tree, briefly.

Stay on Coda if:

Your team has a Doc Maker — one or two people who genuinely enjoy building docs that act like apps — and the rest of the team uses what they build.
Your work has natural table shapes (CRM, content calendar, OKR tracker, applicant pipeline) that benefit from being embedded in a doc.
You use a handful of Packs (Slack, Salesforce, GitHub) that do real work pulling external data into your docs.
You are comfortable with the Superhuman corporate story and prefer Coda’s deliberate update cadence to the more frenetic competition.

Consider leaving if:

Your team uses Coda as a wiki — runbooks, decision records, postmortems, onboarding docs — and the Doc Maker bottleneck means pages get written by exactly one person.
Your search experience consistently fails on queries you know are in the workspace.
Your team has grown past the one Doc Maker can know all the docs size — usually around 25-40 people — and the formula surface has stopped scaling.
You are paying for Coda AI and finding that the AI-on-every-plan competitors (including Raccoon Page via MCP, see the AI knowledge base post) deliver more for the same price.
Strategic uncertainty about the Superhuman roadmap matters to your procurement story.

The signal we’d watch for, more than any of the above: are your team’s most-edited documents owned by exactly one person? A team where the Coda Doc Maker is on holiday and the docs all drift into staleness is a team that has confused a custom app with a shared wiki.

For the migration side of the question — Coda docs round-trip poorly, but a structured Coda export plus Confluence or Notion as a staging surface plus the ten-minute Confluence import or Notion import into Raccoon Page is a workable two- step path. Detail on the trade-offs in the Notion alternatives and Obsidian vs Notion write-ups.

Things people actually ask

What does Coda mean? Coda is the musical term for a passage that brings a piece of music to a close. The founders named the product after it because the original pitch was the doc that brings the spreadsheet, the database, and the app into one place. It is not an acronym.

What is Coda used for? Building docs that behave like small apps — CRMs, project trackers, OKR dashboards, content calendars, applicant pipelines, internal tools that don’t quite need their own software. The full Coda use case is replace four tools with one carefully-built doc, and the carefully-built qualifier is the load-bearing one.

Is Coda free? Yes, on a generous free tier — unlimited free docs, with limits on doc size, row counts, and the absence of Coda AI. Paid tiers (Pro, Team, Enterprise) add AI, larger doc limits, and team features. Coda’s pricing page is the authoritative source — the numbers move enough that we’re not quoting them here.

Is Coda the same as Notion? Adjacent, not the same. Both are workspace tools with block-based editors and database features. The shapes differ: Notion’s mental model is pages-that-are-also-databases; Coda’s is docs-that-are-also-apps-via-formulas-and-Packs. Notion leans wiki-first; Coda leans doc-with-app-superpowers first. See our Notion explainer for the comparison.

What’s a Coda Pack? A Pack is a pre-built integration that connects Coda to an external service. A Slack Pack lets a doc send Slack messages or read channels; a Salesforce Pack syncs rows with Salesforce records; a GitHub Pack pulls issue lists into a table. There are 450+ Packs as of early 2026, both first-party and community-built.

What is Coda AI? The in-product AI assistant. Summarisation, content drafting, formula suggestion, table-aware queries — the standard 2026 doc-AI feature set, with extra emphasis on formula generation. Gated to paid plans with credit limits.

Does Coda replace a wiki? Not really. A Coda doc can be used as a wiki page, but the shape isn’t the same. Wikis are optimised for many writers, many readers, durable artefacts; Coda docs are optimised for one or two Doc Makers building small apps. Most teams that adopt Coda either pair it with a separate wiki or end up re-deriving wiki features inside one giant doc, which is a recognisable anti-pattern.

Who owns Coda now? Superhuman, which is the rebrand of Grammarly, which acquired Coda in December 2024. Coda is one of three products inside Superhuman alongside the Grammarly writing tools and the Superhuman email client.

How do I migrate off Coda? Coda exports to Word, PDF, and CSV-per-table. The doc-as-app structure doesn’t round-trip cleanly — formulas, buttons, Packs, and view configurations are lost on export. The practical migration path is to export the content and the table data, treat the rest as redesign work, and rebuild in the target tool. For wiki-shaped Coda docs, the Notion importer is a reasonable two-step staging path if you don’t want to re-author straight from CSV.

Coda is the doc-and-database hybrid that a small set of teams have built remarkable things on, and that a larger set of teams have quietly outgrown. The honest question is which set you’re in: do you have a Doc Maker, and is the docs-as- apps shape the right answer for your team’s work? If yes, Coda is the tool. If your team is writing runbooks, decision records, and onboarding docs and finding that exactly one person knows where everything lives, the wiki you wanted is a different shape. Raccoon Page Free is three users, one space, one hundred pages, no card — enough to find out.