Cognee pricing, status and price history
Alive and building. Pricing settled on a pure usage meter — Free / Standard at $2.50 per 1M tokens plus $5 per additional workspace / Enterprise — renamed on 2026-07-01 [OBS] and verified live on 2026-07-16 [FDR]. GitHub stars ran 26,391 to 27,949 between 07-02 and 07-16, roughly 111 per day [OBS], against a $7.5M seed announced 2026-02-19 [FDR].
Open-source AI memory that turns data into knowledge graphs
Price history — observed daily
OUR OWN INSTRUMENT, READ EVERY MORNING AT 02:00 UTC SINCE 2026-06-30. THIS TABLE CANNOT BE BACKFILLED — THAT IS THE POINT.
| First read | Entry price | Source |
|---|---|---|
| 2026-06-30 | $2.50 (usage floor) | pricing page [OBS] |
| 2026-07-04 | UNDER REVIEW | Reading failed our sanity gate — quarantined, logged in corrections. We show you the gap, not a guess. |
| 2026-07-16 | $5 (usage floor) | pricing page [OBS] |
The change log — what actually moved
ONLY DAYS WITH CHANGES APPEAR. QUIET DAYS ARE DATA TOO — THEY JUST DON’T GET A ROW.
- 2026-07-24Sitemap grew 136 → 147 pages (+11). [OBS]
- 2026-07-21GitHub stars +300 in a day (28,508 → 28,808). [OBS]
- 2026-07-19GitHub stars +302 in a day (27,994 → 28,296). [OBS]
- 2026-07-16Live check confirmed the model on the page: Free ($0, 1M tokens bundled) / Standard ($2.50 per 1M tokens plus $5 per additional workspace) / Enterprise (contact); site badge showed 28k stars, instrument logged 27,949. [FDR] source
- 2026-07-15Sitemap added one URL (total 134): a post announcing Cognee's selection for the Berkeley Xcelerator 2026 Spring cohort; stars read 27,875. [OBS]
- 2026-07-13Stars logged at 27,650, up 1,259 in 11 days from the 07-02 reading. [OBS]
EXPOSURE: Feeds CALL-0001 — 28,500 stars by Jul 27 — the desk's first live grading of a memory-layer sprint. - 2026-07-07Homepage API verbs were rewritten from cognee.add / cognee.search to cognee.remember / cognee.recall, and the 'Open-source core' label was shortened to 'Open-source'; stars read 27,274. [OBS]
- 2026-07-05A Bayer case study ('From an AI chatbot to agentic research memory at Bayer') replaced University of Wyoming in the featured homepage slot; the two alternated on subsequent readings (07-07, 07-10, 07-13, 07-15). [OBS]
- 2026-07-04A Cognee Cloud block was added to the pricing page ('Start free. Pay one flat rate for the tokens your memory processes'); the block was verified still live on 2026-07-16. [OBS]
- 2026-07-04GitHub stars +424 in a day (26,391 → 26,815). [OBS]
- 2026-07-02Sitemap went from 0 to 129 URLs in one day, exposing an academy course series and case-study pages including Dynamo, Knowunity, University of Wyoming, and a tier-1 US bank. [OBS]
- 2026-07-02First parsed GitHub star reading logged at 26,391. [OBS]
- 2026-07-01Tiers were renamed at unchanged rates: Hobby became Free (1 workspace, 1M tokens included, no card required) and Growth folded into Standard at $2.50 per 1M tokens plus $5 per additional workspace. [OBS]
- 2026-06-30Coverage opened: homepage rendered 7,110 words, sitemap exposed 0 pages, and pricing listed Hobby (usage at $2.50 per 1M tokens) and Growth ($5 per workspace per month). [OBS]
The clerk’s read
Cognee's cloud pricing is on the books as a pure metering play. The 07-01 rename — Hobby to Free, Growth folded into Standard — kept every number and moved only the frame: $2.50 per 1M tokens processed, $5 per additional workspace, no seat pricing anywhere on the page. Verified live on 07-16: Free carries 1M bundled tokens with no card, Standard is usage plus a workspace add-on, Enterprise is contact-only. [INF] The unit choice fits the product's shape: a memory layer's cost driver is tokens cognified and recalled, not humans logged in, and metering the token also prices the open-source escape hatch honestly — a self-hoster's alternative cost is compute, not licenses. The 07-04 pricing-page addition spelled the pitch out: 'Start free. Pay one flat rate for the tokens your memory processes.' Around the pricing work, the funnel got built fast: sitemap 0 to 129 URLs on 07-02, 134 by 07-15, stocked with academy pages and case studies (Bayer, Knowunity, a tier-1 US bank). Stars ran 26,391 to 27,949 between 07-02 and 07-16 — roughly 111 per day, with the homepage badge ticking 26.4k to 28k over the same window (26.1k on 07-01). CALL-0001 stands open on the books: 28,500 stars by Jul 27. The 07-16 reading leaves 551 to cover. Filed and carried forward.
The watch line on this board
Is the star sprint momentum or a spike?
NEXT SIGNAL: CALL-0001, open on this board: 28,500 stars by Jul 27, graded the next morning against the deadline-day reading.
STOPS MATTERING IF: A post-verdict stall — velocity halving from the ~111/day pace filed across 07-02 to 07-16 — would close the momentum read regardless of the stamp.
A READ WITHOUT A NEXT SIGNAL IS JUST AN OPINION — THIS ONE IS FILED IN ADVANCE, LIKE EVERYTHING ELSE ON THE DESK.
The open call on this board
NO OPEN CALL ON THIS BOARD — THE DESK ONLY STAKES A CLAIM WHEN THE EVIDENCE MOVES. LAST GRADED 2026-07-28.
Graded calls on this board
Cognee crosses 28,500 GitHub stars by Jul 27.
Why it matters: The memory-layer race runs on funding and star velocity, not invoices. Cognee has led the track's weekly star gains since we started counting — this call tests whether the sprint is momentum or a spike.
The facts on file
| Traction (verified) | $7.5M raised (2026-02) · 12k+ stars [3P] source |
|---|---|
| Team | Berlin team |
| Launched | 2024 |
| On the books since | 2026-06-30 |
| Entry price | $5 (usage floor) · 1 plans pricing page · read 2026-08-29 · other readings under review |
| Sitemap pages | 161 sitemap.xml · read 2026-08-29 |
| GitHub stars | 30328 public GitHub count · read 2026-08-29 |
Instrument readings link to their target and carry the date we read them; roster context (team, launch) is filed at intake. Where we can’t verify, the cell says UNKNOWN — never a guess. Dispute a number: /corrections/.
The field — AI Memory Layer, 7 on the books
| Board | Role | Entry price | Status |
|---|---|---|---|
| BasicMemory | — | $12/mo | WATCHING |
| Cognee — this board | — | $5 (usage floor) | PROVEN |
| Goldfish | — | — | WATCHING |
| Letta | — | $20/mo | PROVEN |
| Mem0 | — | $19/mo | PROVEN |
| Supermemory | — | $5/mo | PROVEN |
| Unabyss | — | $13/mo | WATCHING |
FULL FIELD COMPARISON: AI MEMORY LAYER →
Buyer’s note — the facts if you’re paying
Cognee prices on a pure meter, verified live 07-16: Free (1M tokens bundled), Standard at $2.50 per 1M tokens plus $5 per additional workspace, Enterprise by contact. There is no seat tier to outgrow — the bill scales with tokens, so how it compares with the track's flat $12–$19 subscriptions depends entirely on your volume. [FDR 2026-07-16]
STATED FROM THE TAPE: WHAT CHANGED AND WHO IT PRICES IN OR OUT. NO ADVICE, NO IMPERATIVES. READINGS AND VERDICTS ARE NEVER AFFECTED BY AFFILIATE LINKS.
TRY COGNEE →PLAIN LINK, NO AFFILIATE TAG TODAY. IF THAT EVER CHANGES, THIS LINE WILL SAY SO AND THE LINK WILL CARRY rel="sponsored". THE VERDICTS ABOVE DON’T CARE EITHER WAY.
Sources on this page
- [FDR] Cognee pricing page (fetched live 2026-07-16) — https://www.cognee.ai/pricing
- [FDR] Cognee homepage (fetched live 2026-07-16) — https://www.cognee.ai
- [FDR] Cognee seed announcement, 2026-02-19 — https://www.cognee.ai/blog/cognee-news/cognee-raises-seven-million-five-hundre…
- [OBS] Radar change log r025 (daily instrument, 2026-06-30 to 2026-07-15) — https://www.cognee.ai
- [3P] Traction figure — https://www.eu-startups.com/2026/02/german-ai-infrastructure-startup-cognee-la…
- [OBS] Daily instrument readings (homepage, pricing page, sitemap.xml, GitHub) — read at 02:00 UTC, on the books since 2026-06-30
Dispute a number? /corrections/ — handled in public within 48 hours.