Memory API for agents

Long-term memoryfor AI agents thatnever forget a user

Give every agent a memory it keeps across sessions, users and deploys. Write with one call, recall in under 40 ms, and decide what fades and what stays.

Start buildingStart building Read the docsRead the docs

HaldenQuorraTallylineOspreyWavefoldNorthstarBrightseatKestrelHaldenQuorraTallylineOspreyWavefoldNorthstarBrightseatKestrel

Hold to write

wrote “Prefers aisle seats” · mem_8f2k · 41 ms

Production memory without the production plumbing.

Context that survives every session

Why Plivel

01 / 04

recall(user="dana_k", q=")

MemoryRelevance

Prefers aisle seats, never exit rows0.00

pinned

Flies TAP from Newark, Star Alliance Gold0.00

3 recalls

Asked for a vegetarian meal in May0.00

decaying

3 of 12,408 memories0 ms

The right memory, before the model starts talking.

Hybrid vector and graph search runs next to your agent in 14 regions. Voice agents get context inside the first breath, not after an awkward pause.

Pinned · penicillin allergyPreference · aisle seatOne-off · parked at SFO lot C

Memories fade unless they earn their keep.

Every memory carries a relevance score that decays over time and climbs each time it is recalled. Pin what must never fade, like an allergy, and let one-off chatter archive itself.

agent.py

written mem_8f2k · merged with 2 · 41 ms

Bring your agent. Keep your stack.

Python, TypeScript or plain HTTP. No new framework, no schema to design first. Write what the user said, recall what matters, and Plivel handles embedding, dedupe and merge.

TRACE tr_91ce · recall(dana_k)0ms

auth + tenant

1.2 ms

embed query

6.1 ms

vector search · 14 shards

14.8 ms

graph expand · 2 hops

7.3 ms

rerank · top 40 → 3

6.0 ms

compose context

2.6 ms

REGION FRA-2 · 14 SHARDSEXPORT → OTEL

See exactly why your agent remembered that.

Each recall ships a span-level trace: embed, search, graph expand, rerank. Export to OpenTelemetry, or debug in the Plivel console with the prompt it fed the model.

Life of a memory

One sentence, from the first word to cold storage.

Write

Your agent sends what the user said. One call, no schema.

Embed

Plivel turns it into a 1,536-dimension vector and strips anything marked private.

Link

The memory joins the user's graph, next to the facts it relates to.

Recall

Weeks later, a new question pulls it back in 38 ms.

Fade

Unused, its relevance decays. Every recall tops it back up.

Memory that lives next to your agent

Each user's memory shard is placed in the region closest to where your agent runs, and pinned there if the law says it has to stay. Failover is automatic.

Regions

0 ms

p50 recall

0.00%

Uptime SLA

MEMORY SHARDS · 14 REGIONS

FRANKFURT · P50 RECALL 25 MSDATA RESIDENCY: EU · US · APAC PINNABLE

Security

Private by default, erasable on request

Trust center ↗Security docs

AES-256 AT REST · PER-TENANT KEYS · TLS 1.3

SOC 2 Type II, GDPR, HIPAA

Audited yearly. Reports and our BAA are one click away in the dashboard.

Residency you can pin

Keep EU users' memories in Frankfurt or Stockholm, and prove it with region-locked keys.

Tenant isolation

Every customer gets separate encryption keys and shards. No shared indexes, ever.

Erase with a receipt

One call deletes a user everywhere within 60 seconds and returns a signed receipt.

Built with Plivel

Recipes for agents that know their people

Receptionist that knows the regularsGreets returning callers by name and skips the questions it already has answers to.Open recipe →

Sub-second context for LiveKitRecall runs while the caller is still speaking, so the first reply is already personal.Open recipe →

Outbound follow-upsCalls back with the notes from last time, not a blank script.Open recipe →

Language it heard last timePicks up in Portuguese if that is how the caller ended the previous call.Open recipe →

Benchmarks

Faster, cheaper, and it remembers the right thing

Measured September 2026 on 10M synthetic memories across 40k users, same embedding model for every setup, recall from fra-2. Scripts and raw numbers are in the docs so you can rerun them.

p95 recall latency at 10M memories · lower is better

Plivel

0 ms

Vector DB + Postgres glue

0 ms

Re-summarise history with an LLM

0 ms

■ Customer stories

Teams that stopped starting from zero

All stories →

Healthcare · Voice

Halden Health

How Halden's nurse line stopped asking patients the same questions twice

-38%

average call length

Fintech · Support

Tallyline

Tallyline cut repeat support tickets by a third with an agent that remembers the last fix

-33%

repeat tickets

Developer tools · Coding

Wavefold

Wavefold's coding agent learned each team's conventions once, instead of every session

4.1x

fewer clarifying questions

Travel · Commerce

Brightseat

Brightseat's booking assistant remembers aisle seats, cellos and frequent-flyer numbers

+19%

repeat bookings

Education · Tutoring

Osprey Learning

Osprey's tutors pick up exactly where each student left off

+27%

lesson completion

Legal · Research

Quorra Legal

Quorra keeps every client's matter history in memory, and erases it on request

60 s

to erase a client everywhere

Utilities · Voice

Northstar Energy

Northstar's outage line remembers who is on medical equipment

38 ms

p50 recall during peak storms

Robotics · Field ops

Kestrel Robotics

Kestrel's field robots remember every site they have mapped and every operator's preferences

-52%

re-mapping time per site

Transparent pricing

Price it on your real workload

Move the sliders to match your agent. You pay for writes, recalls and storage, and nothing for seats, agents or users.

Estimated monthly cost

$301PRO PLAN

$0.019 per active user · a DIY vector-DB setup would run about $2,456

Writes$228

Recalls$23

Storage$1.43

Plan$49

2,282,256 memories stored5,705,640 recalls / mo

See all plansSee all plans

Monthly active usersPeople your agent talks to15,849

Memories written per userPer month, after dedupe24

Recalls per userPer day12

Keep before archivingMonths, unless pinned6 mo

Regions

Writes $0.60 per 1k incl. embedding · Recalls $4 per 1M · Storage $0.25 per GB-month (~400k memories) · Extra regions +15%

■ Changelog

Shipped every other Thursday

Full changelog →

Sep 18, 2026

v4.12

Recall traces in the console

Every recall now opens as a span waterfall with the exact context your agent received.

→

Feature

Sep 4, 2026

v4.11

Pinned memories never decay

Mark a memory as pinned and it keeps full relevance forever. Built for allergies, contracts and consent.

→

Feature

Aug 21, 2026

v4.10

São Paulo and Johannesburg regions

Two new regions bring the total to 14. Median recall from Latin America drops from 118 ms to 36 ms.

→

Improvement

Aug 7, 2026

v4.9

Faster graph expand

Two-hop graph expansion is 2.3x faster at p95 after we moved edge lists next to the vectors.

→

Improvement

Documentation

Docs written by the people who answer the pager

Every endpoint has a runnable example, every guide ends with a working agent. Press ⌘K and try “decay”.

⌘K

Quickstart in 5 minutesdocs / get-started / quickstart-in-5-minutes→Install the Python SDKdocs / get-started / install-the-python-sdk→Install the TypeScript SDKdocs / get-started / install-the-typescript-sdk→Authentication and keysdocs / get-started / authentication-and-keys→

Questions

Asked on every sales call

No. A vector store is one part. Plivel adds per-user isolation, a memory graph, decay and pinning, merge and dedupe, traces, and erase-by-user for GDPR. You would otherwise build those around the database yourself.

Any. Plivel returns plain text context and structured memories. It works with OpenAI, Anthropic, Mistral, Llama or your own fine-tune.

In the region you choose, encrypted at rest with per-tenant keys. EU-only and US-only residency are one setting. We never train on your data.

One call: DELETE /v1/users/:id. Memories, vectors and graph edges are erased within 60 seconds, and you get a signed receipt for your records.

Recall has a 99.99% SLA and fails over across regions. Agents can also run with a local read cache that serves the last known context.