Mem0 pricing

About

Mem0 is a drop-in memory layer for AI agents and applications — infrastructure that lets an agent remember facts about users and prior interactions across sessions so context persists in production. Developers send “memories” to Mem0 and later retrieve the relevant ones at inference time, offloading long-term memory from the prompt window to a managed store. Founded in 2023 by Taranjeet Singh and Deshraj Yadav (YC S24), the company pivoted from Embedchain, their earlier open-source RAG framework, after users of a viral meditation app they built kept complaining that it couldn’t remember them between sessions.

Mem0 ships in two forms: an open-source library (Apache 2.0, 41,000+ GitHub stars) that teams can self-host for free, and a hosted Platform with managed storage, analytics, and support — this page covers the hosted Platform’s pricing. In October 2025 Mem0 announced $24M in funding ($3.9M seed led by Kindred Ventures plus a $20M Series A led by Basis Set Ventures, with Peak XV Partners, GitHub Fund, and Y Combinator participating), reporting API calls growing from 35 million in Q1 to 186 million in Q3 2025 and a slot as the exclusive memory provider for AWS’s Agent SDK. It positions against other agent-memory products such as Zep, Letta, Cognee, Mastra, Honcho, Hindsight, and Supermemory — all listed in its own comparison footer.

The hosted Platform is sold as four flat monthly tiers plus a custom Enterprise plan, all metered on memory operations rather than seats. For the current live pricing, see Mem0’s pricing page.

Pricing summary : quota-gated subscription metered on memory operations

Mem0’s hosted Platform uses a flat-tier subscription with a freemium entry, where each tier unlocks a larger monthly quota across two metered dimensions:

1. Add requests (memory writes): the count of operations that store a new memory each month — 10,000 (Hobby) → 50,000 (Starter) → 200,000 (Growth) → 500,000 (Pro), unlimited on Enterprise.
2. Retrieval requests (memory reads): the count of search/recall operations each month — 1,000 → 5,000 → 20,000 → 50,000, unlimited on Enterprise.
3. Projects: workspace count — 1 → 1 → 3 → unlimited (Pro and Enterprise). Support and analytics also step up by tier (community → email → private Slack → Slack + SLA; none → basic → advanced).

The four self-serve tiers are fixed monthly fees (Free, $19, $79, $249); Enterprise is a custom, sales-led quote and is where Mem0 offers explicit usage-based pricing for teams that outgrow the fixed quotas — a dedicated callout on the pricing page (“We also support usage-based pricing”) routes them to sales.

What makes this different: Mem0 bills on memory operations, not seats — end users are unlimited on every tier, including the free Hobby plan. The price you pay scales with how much your agents read and write to memory, decoupled entirely from how many people use the app. And because the core library is Apache 2.0, the hosted tiers sell convenience, analytics, and support rather than the capability itself.

Pricing by product

Hosted Platform (self-serve plans)

Hosted Platform (Enterprise)

The public pricing table leaves the on-prem deployment, audit logs, custom integrations, and SSO rows blank across all listed tiers, so those capabilities are not separately priced on the page; Enterprise terms are set during the sales conversation via the Mem0 enterprise contact form.

Open source (self-host)

The core Mem0 library is free under Apache 2.0 with no operation caps — you run your own vector store and pay your own LLM costs for the extraction and retrieval calls Mem0 makes under the hood. The paid Platform adds managed hosting, the analytics dashboard, and support.

Startup program

Startups with under $5M in funding can apply for 3 months of free access to the Pro plan (advertised as worth $1,000), with priority support and direct collaboration with the Mem0 team. The grant was twice as generous as recently as September 2025 (see Pricing evolution).

Sales motions across products: PLG / self-serve for Hobby, Starter, Growth, and Pro (plus the OSS funnel); sales-led for Enterprise and its usage-based option.

Hidden costs : What Mem0 users actually pay

Mem0’s self-serve pricing is unusually clean — flat fees, two published quotas, no per-seat math. The catch is what the page does not say: there are no published overage rates, so the real cost question is what happens when an agent workload crosses a quota mid-month (upgrade, or a sales conversation for usage-based pricing). And retrieval quotas are 10x smaller than add quotas on every tier, so read-heavy agents hit the ceiling far sooner than the headline add number suggests.

Two structural costs to budget for: the retrieval ratio (1:10 against adds on every tier) effectively makes retrieval the binding constraint for conversational agents that recall context on every turn; and analytics gating means you can’t see detailed usage patterns on Hobby or Starter — the visibility that helps you predict an upgrade only arrives once you’ve already upgraded to Growth.

Want to estimate your own Mem0 bill? Use the Mem0 pricing calculator to model your costs based on usage patterns.

Pricing evolution : Mem0 pricing history and changes

Cadence

Tracked range: 2024–present, via Wayback Machine snapshots (2024-08, 2024-10, 2025-03, 2025-09, 2025-11, 2026-03, 2026-04, 2026-05) and a live 2026-06-10 capture.

Notable changes

• 2024-08 — Earliest archived pricing: Open Source (free, unlimited), “Pro – Hosted” with 10k API requests free/month and no published paid rate, and a custom-limits Enterprise. Classic OSS-first launch posture.
• 2024-10 — Restructures around the hosted Platform with memories as the value metric: Free (10,000 memories, 1,000 retrieval calls/mo, unlimited end users) / Pro / Enterprise.
• 2025-03 — Pro’s $249/month price first appears in the archive; the free tier is renamed Hobby; Enterprise is labeled “Flexible Pricing”.
• 2025-09 — Starter added at $19/month (50,000 memories, 5,000 retrievals), creating a low-commitment paid step. Startup program offers 6 months of Pro free (worth $1,500).
• 2025-11 — Startup program quietly trimmed to 3 months (worth $1,000), around the same time the pricing page picked up a banner for the $24M Series A (announced 2025-10-28).
• 2026-04 — Value metric renamed from “memories” to “add requests” and “retrieval requests” — same quotas, more precise operation language.
• 2026-05 — Growth tier added at $79/month (200,000 add, 20,000 retrieval, 3 projects, email support, basic analytics), plugging the 13x jump between $19 and $249.
• 2026-06-10 — Live capture confirms the five-plan structure: Free / $19 / $79 / $249 / custom Enterprise, plus the usage-based callout and 3-month startup grant.

What’s unique : Mem0’s distinctive pricing mechanics

1. A two-sided meter: writes and reads priced as separate quotas. Most API products meter one thing (requests, tokens, credits). Mem0 splits its value metric into add requests (writing memories) and retrieval requests (reading them), each with its own quota per tier — and holds the ratio at roughly 10:1 adds-to-retrievals on every plan. That mirrors the underlying cost asymmetry (writes trigger LLM extraction; reads hit a vector index) and lets Mem0 price the two behaviors independently.

2. Unlimited end users on every tier, including free. Mem0 explicitly lists “END USERS: UNLIMITED” as the first row of its comparison table on all five plans. For a B2B2C infrastructure product this is the whole pitch: your app can have a million users, and you pay only for the memory operations they generate. There is no seat dimension anywhere in the pricing.

3. The meter rename: “memories” → “add requests”. In April 2026 Mem0 renamed its units from fuzzy nouns (“10,000 memories”) to precise verbs (“10,000 add requests”) without changing the quotas. It’s a subtle transparency upgrade — “memories” implied a stored-object count (stock), while “add requests” correctly describes a metered operation (flow), which is what was actually being counted all along.

Strengths & weaknesses

Billing UX : self-serve sign-up, usage-based escape hatch, and a qualified enterprise form

• Per-tier CTAs — each plan card carries its own call to action: “Start for Free” (Hobby), “Get Started” (Starter, Growth, Pro), and “Contact Us” (Enterprise), so paid tiers begin self-serve while Enterprise routes to sales.
• “Talk to our team” usage-based path — a dedicated “Usage based pricing” callout below the tier table (“Not finding something that fits your need? We also support usage-based pricing.”) routes teams who exceed the fixed quotas to a custom usage-based quote.
• Enterprise qualification form (app.mem0.ai/enterprise) — a lead form that collects work email, company size (1-10 through 5,000+), use case (AI companion, AI Healthcare, AI Customer Support, AI Agent Infra), decision timeline, and the reason fixed pricing doesn’t fit, before producing a quote.
• Startup Program application — a self-serve apply flow (“Apply for Startup Program”) grants qualifying startups (under $5M raised) 3 months of the Pro plan free.
• Tiered analytics — analytics is itself a billing-tier control: none on Hobby/Starter, “basic” on Growth, and “advanced” on Pro and Enterprise, giving usage visibility only on the higher plans.

Strategic wins : Why Mem0’s pricing decisions worked

1. Pricing the operation, not the seat or the user

Memory is consumed by agents, not humans, so Mem0 metered the agent behavior (add/retrieval operations) and made end users unlimited everywhere. That removed the worst adoption blocker for B2B2C builders — per-user infrastructure pricing that scales with their customer count — and aligned Mem0’s revenue with actual workload. The bet is validated by the volume curve: API calls grew from 35M to 186M per quarter during 2025. See choosing the right usage metric.

2. OSS as the funnel, Platform as the monetization

Like the strongest AI-infra playbooks, Mem0 leads with an Apache 2.0 library (41,000+ stars, 14M downloads) and monetizes hosting, analytics, and support. The OSS footprint earned it default-integration status in frameworks like CrewAI, Flowise, and Langflow — and the AWS Agent SDK exclusive — which the hosted tiers then convert. See how AI companies structure pricing.

3. Laddering the gap as the market matured

The paid ladder was built incrementally and in response to demand: a single $249 Pro (early 2025), then a $19 Starter to catch indie builders (late 2025), then a $79 Growth to plug the 13x jump in the middle (May 2026). Each insertion lowered the upgrade cliff rather than repricing existing customers — quotas and prices already published never moved. Related: outcome-based pricing trends.

Areas to improve : Gaps in Mem0’s pricing approach

1. Publish what happens at the quota edge

The page lists quotas but never says what happens when you cross one — hard stop, throttle, or forced upgrade — and publishes no overage rates. For an agent workload that can spike 10x overnight, that’s the single most important unanswered billing question. See bill shock and cost unpredictability.

2. Put numbers on the usage-based option

“We also support usage-based pricing” is an escape hatch without a price. Publishing even indicative per-1,000-operation rates would let scaling teams model costs past the $249 tier instead of entering a quote process blind — and would make the Enterprise conversation faster, not slower.

3. Rebalance (or explain) the retrieval ratio

Every tier holds retrievals to a tenth of adds, but memory only creates value when it’s read. Conversational agents that recall context on every turn will exhaust retrieval quotas long before add quotas, making the effective price-per-useful-operation much higher than the headline suggests. Either rebalancing the ratio or explaining the cost logic behind it would build trust with exactly the heavy users Mem0 wants.

Key takeaways

1. Meter the agent’s behavior, not the human’s seat. Mem0 charges for memory operations and gives away unlimited end users — the right shape for infrastructure consumed by software, not people.
2. A split read/write meter can mirror your cost structure. Separate add and retrieval quotas let Mem0 price LLM-heavy writes and index-hit reads independently — but the ratio you choose becomes a hidden constraint for customers.
3. Build the ladder one rung at a time. $249 first, then $19 below it, then $79 in the middle — Mem0 added tiers as segments appeared instead of guessing the full ladder upfront, and never repriced a published tier.
4. Rename your meter toward precision. “Memories” → “add requests” cost nothing and removed ambiguity about whether customers pay for stored stock or operation flow.
5. An unpublished escape hatch is half a pricing model. The usage-based option exists only as a sales callout; the quota-edge behavior and unit rates are the corpus-wide transparency gap to watch.

UBP implications

1. Quota-gated tiers are the bridge between subscription and pure usage. Mem0 sells predictable flat fees while the meters underneath (add/retrieval operations) do the segmentation — usage-based pricing’s alignment without its bill anxiety. See usage-based pricing strategy.
2. Pick the unit your buyer’s code already emits. Developers instrument add() and search() calls anyway; quotas denominated in those exact operations require no new mental model — the meter is the API.
3. When OSS caps your pricing power, meter the managed value. With the core library free forever, Mem0’s quotas price exactly what self-hosting costs you in toil: managed storage, analytics, and support around the same operations.

Sources

• Mem0 pricing page (live capture, accessed 2026-06-10)
• Mem0 enterprise contact form (live capture, accessed 2026-06-10)
• Mem0 — $24M Series A announcement (accessed 2026-06-10)
• PR Newswire: Mem0 raises $24M Series A (accessed 2026-06-10)
• TechCrunch: Mem0 raises $24M from YC, Peak XV and Basis Set (accessed 2026-06-10)
• Mem0 GitHub (Apache 2.0) (accessed 2026-06-10)
• Y Combinator — Mem0 (S24) (accessed 2026-06-10)
• Wayback Machine snapshots: mem0.ai/pricing — 2024-08, 2024-10, 2025-03, 2025-09, 2025-11, 2026-03, 2026-04, 2026-05 (accessed 2026-06-10)

Bottom line

Mem0 (YC S24) turned an open-source memory library into the default memory layer for AI agents — Apache 2.0 core, 41,000+ GitHub stars, AWS Agent SDK exclusivity, and $24M in funding as of October 2025. Its hosted pricing is a quota-gated subscription metered on memory operations: free Hobby, $19 Starter, $79 Growth, and $249 Pro, each with separate monthly add-request and retrieval-request quotas and unlimited end users, plus a custom usage-based Enterprise. The ladder was built rung by rung (Pro first, Starter and Growth later) without ever repricing a published tier; the open questions are unpublished overage behavior and a 10:1 add-to-retrieval ratio that squeezes read-heavy agents. Browse the pricing blueprint for more fully-researched company profiles.

Want to compare Mem0 against other AI agent infrastructure companies? Browse the pricing blueprint.

Pricing timeline : Major events on a vertical axis

Each milestone below corresponds to a public pricing change, product launch, or material adjustment. Major events use a filled marker; minor adjustments use a faded one.

Five-plan structure verified: Free / $19 / $79 / $249 + custom Enterprise

Jun 2026

Live capture confirms Hobby (Free), Starter ($19/mo), Growth ($79/mo), Pro ($249/mo), and custom Enterprise, all metered on monthly add and retrieval requests with unlimited end users, plus a usage-based pricing callout and the 3-month Pro startup program.

Growth $79 tier added; meters renamed add / retrieval requests

May 2026

Mem0 renames its meters from 'memories' to 'add requests' and 'retrieval requests' (by April 2026) and inserts a $79/month Growth tier (200,000 add, 20,000 retrieval, 3 projects, email support, basic analytics), plugging the 13x price jump between Starter and Pro.

Starter $19/month added + startup program (6 months of Pro, $1,500 value)

Sep 2025

A $19/month Starter tier (50,000 memories, 5,000 retrieval calls) slots between Hobby and the $249 Pro. A startup program offers companies under $5M in funding 6 months of Pro free, advertised as worth $1,500 — later trimmed to 3 months / $1,000 by November 2025.

Pro lands at $249/month; free tier renamed Hobby

Mar 2025

First archived snapshot showing a published Pro price: $249/month with private Slack support, advanced analytics, and multi-project support. The free tier is renamed Hobby (10,000 memories, 1,000 retrievals) and Enterprise is labeled 'Flexible Pricing'.

Platform tiers metered on memories: Free / Pro / Enterprise

Oct 2024

Pricing reorganized around the hosted Platform with memory operations as the value metric: a Free tier with 10,000 memories and 1,000 retrieval API calls per month plus unlimited end users, a Pro tier, and a custom Enterprise plan.

Launch-era hosted pricing: OSS free + 'Pro – Hosted' with 10k free requests

Aug 2024

Earliest archived pricing page shows three cards: Open Source (free and unlimited, self-managed), 'Pro – Hosted' with 10k API requests free per month ('Start now for free' — no paid rate published), and a custom-limits Enterprise – Hosted plan.