memgramBeta
Start Free
memgram · now in public beta

Memory for AI agents you can actually inspect.

As conversations grow, AI agents start making memory decisions you can't see. memgram makes retrieval, persistence, and memory state inspectable in production.

Start Free
Backed bytechstars_NYC
app.memgraph.com — memory layer
live
conversation
I usually work late at night and prefer step-by-step troubleshooting.
Noted — I'll keep responses detailed and sequential.
When can you get my MacBook error sorted?
Let's walk through it step by step, given your late-night setup.
memory pipeline
124ms
1
extract
ok
Prefers step-by-step troubleshooting.
2
classify
ok
preference · support style
3
deduplicate
ok
no similar memory — new
4
decide
persisted
persist · scope=support.shared
memory state
demo_0763f3333 active
preference0.92
Prefers step-by-step troubleshooting over quick replies.
fact0.78
Usually works late at night.
constraint0.71
Frustrated by repeating the same issue across conversations.
Scope: support.shared
events
2,418
persisted
1,892
rejected
526
p50 trace
118ms
VPC deployableOpen core · coming soonOpenTelemetry compatible
01 · Problem

AI memory becomes unreliable at scale.

Short conversations don't need memory systems. But once sessions get long, context fades, retrieval gets noisy, agents drift, and developers lose visibility into why.

stage 1
Conversation
long-running
stage 2
Memory extraction
what's worth keeping?
failure
stage 3
Retrieval
what's relevant now?
failure
stage 4
AI response
consistent with state?
context fades across turns
retrieval becomes inconsistent
memory gets polluted across agents
no answer to: why did it remember that?
02 · Why current memory systems break

Today's memory is a black box.

Memories get stored automatically, retrieved probabilistically, and you have no real answer to what your system believes or why.

Current systems
memory.store(?)
???
memory.recall(?)
  • stores memories automatically
  • retrieves probabilistically
  • hides extraction logic
  • no observability
memgram
app.memgraph.com — users / alice
live
user
alice
first seen 20h ago · last active 14m ago
myagent1 · 2myagent2 · 1
memory state
3 active
Scope: support.shared — shared across 3 support agents
preference
prefers coffee over tea
0.78
location
lives in New York (since May)
0.92
context
csv export issue, recurring
0.50
activity — newest first
all agents ▾all events ▾
SEARCH3 results · 84ms14m
what is alice experiencing
WRITEmyagent2 · 1 stored20h
User is experiencing a CSV export issue.
indicates recurring problem · context · 0.50
WRITEmyagent1 · 1 rejected20h
User is experiencing a CSV export issue.
duplicate of memory_4f1 · rejected
  • every decision is a traceable event
  • retrieval comes with a reason
  • inspect what the system believes
  • policies you control
03 · Capabilities

The control plane for AI cognition.

Cognition pipeline

See extraction, classification, deduplication, and persistence decisions stream through a typed pipeline in real time.

Retrieval causality

Every retrieval comes with a reason — why this memory surfaced, what was scored, and what was rejected.

Belief state inspector

Open any user, session, or agent to see exactly what the system currently believes — and how it got there.

Policy & governance

Declarative policies for what persists, what expires, what merges, and what stays isolated — per scope.

Multi-agent scopes

Prevent cross-agent contamination. Inspect every state transition between agents sharing the same user.

Managed or self-hosted

Start on managed cloud today. Self-hosted VPC deployment coming soon — same APIs, same control plane.

04 · The product

Inspectable runtime state for AI agents.

Trace every state transition. See what your agent currently believes. Replay any retrieval decision. memgram is the missing developer surface for production AI cognition.

1 persisted

Why this memory was retained.

Every memory event opens into a full trace: the user message, extraction instructions, and every step of processing — extract, classify, deduplicate, decide.

  • step-by-step pipeline replay
  • per-event extraction reasoning
  • reportable, shareable trace IDs
app.memgraph.com — trace 5a565eef
live
Trace
5a565eef…
alice · myagent2 · sess_001… · 20h ago
1 persisted
user message
I'm having the CSV export issue again
I can help with that
extraction instructions
Extract product issues, frustrations, feature requests, and
resolution history. Skip pleasantries and greetings.
memory processing
extract
42mscompleted
"User is experiencing a CSV export issue."
classify
18mscompleted
categorised as · context
deduplicate
22mscompleted
no similar memory found — treating as new
decide
11mscompleted
persist · scope=user_4f1 · confidence 0.62
app.memgraph.com — users / alice
live
user
alice
first seen 20h ago · last active 14m ago
myagent1 · 2myagent2 · 1
memory state
3 active
Scope: support.shared — shared across 3 support agents
preference
prefers coffee over tea
0.78
location
lives in New York (since May)
0.92
context
csv export issue, recurring
0.50
activity — newest first
all agents ▾all events ▾
SEARCH3 results · 84ms14m
what is alice experiencing
WRITEmyagent2 · 1 stored20h
User is experiencing a CSV export issue.
indicates recurring problem · context · 0.50
WRITEmyagent1 · 1 rejected20h
User is experiencing a CSV export issue.
duplicate of memory_4f1 · rejected
Belief state

What the system currently believes.

Open any user, session, or agent to see the active belief state alongside every write, search, and rejection. Catch cross-agent contamination before it reaches production.

Playground

Watch cognition evolve in real time.

Chat with an agent and see — turn by turn — what's extracted, what persists, what's retrieved, and what's rejected. Copy the integration straight into your stack.

Playground — watch memory work in real time
live
chat
I usually work late at night, prefer step-by-step troubleshooting instead of quick replies, and get frustrated when I have to repeat the same issue in every support conversation.
I understand the importance of clear, step-by-step guidance, especially when working late. If you're facing a specific issue, please share the details, and I'll provide a thorough troubleshooting process to help you resolve it efficiently.
no memory context
When can you get my macbook error sorted?
Let's go through the troubleshooting steps for your MacBook error. Please describe the issue you're facing, and we'll work together on resolving it step-by-step.
3 memories used
Type a message…
what's happening
↓ Session started650ms
0 memories for demo_0763f333 · cold start
TURN 1
Search skipped · no memories yet
⚙ Processing complete
extracted
works late at night
prefers step-by-step
frustrated by repetition
TURN 2
search
1182ms
→ 3 results
session memory
User usually works late at night.
factturn 1
User prefers step-by-step troubleshooting instead of quick replies.
preferenceturn 1
User gets frustrated when having to repeat the same issue in every support conversation.
constraintturn 1
retrieved in last turn
0.70User prefers step-by-step troubleshooting.
0.60User gets frustrated repeating issues.
0.50User usually works late at night.
integration code
turn 2
1 · session
import memgram
 
session = memgram.Session(
agent_slug="demo-agent",
user_id="demo_0763f333",
)
 
2 · search memory
# before reply
memories = session.search(
query="When can you get my macbook error…",
)
# → 3 results
 
3 · store async
# after reply
session.add(messages=[…])
# → 0 stored, 0 rejected
05 · How it works

One pipeline. Every decision, visible.

memgram sits between your application and your model runtime. Every memory event passes through a typed pipeline you can introspect and control.

01
AI App
your agent, copilot, or workflow
02
memgram SDK
drop-in for TS, Python, Go
03
Memory processing
extract · classify · dedupe · decide
04
Observable memory layer
state, traces, policies
05
AI runtime
consistent, inspectable behavior
// instrumentation is one line
import { memgraph } from '@memgraph/sdk'
await memgraph.record(conversation)
const ctx = await memgraph.recall({ scope: user.id })
07 · Built for real agents

Templates for the agents you actually ship.

Start from a memory profile tuned to your agent's job — support, personal assistant, coding copilot, sales, health. memgram configures extraction and retention automatically.

app.memgraph.com — new agent
live
new agent · step 1 of 3
What does this agent do?
We'll configure memory extraction automatically.
Support bot
tickets · resolutions · history
Personal assistant
preferences · routines · events
Coding assistant
stack · conventions · repos
Sales bot
leads · intents · objections
Health assistant
vitals · habits · medications
Custom
define your own extraction rules
preview: extracts tickets · resolutions · sentiment
Continue
08 · Production failures

The failures you can't debug today.

Every team running agents in production has hit these. Without an inspectable cognition layer, the root cause looks like a hallucination — but it's a state transition no one saw.

stale context

Customer support agent answered with last quarter's data.

A SaaS team in New York shipped an LLM support agent. Two weeks in, it confidently quoted a deprecated billing flow because that memory was never invalidated.

memgram exposes
  • what was extracted from the old ticket
  • why it persisted past the policy window
  • which retrievals scored it above the new policy doc
cross-agent contamination

Sales agent leaked a support user's frustration into an outbound email.

Two agents shared the same user scope without an isolation policy. The sales agent retrieved a support context memory at the wrong moment — and personalised on it.

memgram exposes
  • which agent wrote which memory
  • why the cross-scope retrieval matched
  • the exact policy rule that should have blocked it
retrieval drift

Coding copilot stopped recommending the team's actual stack.

Over 6,000 turns the embedding space drifted. The copilot started surfacing generic answers because team-specific preferences fell below the retrieval threshold.

memgram exposes
  • confidence decay over time per memory
  • rejected retrievals and their scores
  • state transitions that changed the agent's belief
06 · Deployment

Managed cloud today. Self-hosted soon.

Start on managed cloud with enterprise-grade security. Self-hosted VPC and open-core distribution are landing for design partners — same APIs, same control plane.

Open core · self-hosted · coming soon
api.memgram.io · control plane
managed · p50 118ms
$ curl -X POST https://api.memgram.io/memory/add \
-H "X-API-Key: mem_xxx..." -d '{agent_slug:"support-bot",user_id:"usr_4f1",...}'
→ { trace_id: "3a8f1b2c", status: "processing" }
$ curl https://api.memgram.io/trace/3a8f1b2c
persist · fact · "User lives in Seattle" ✓ new
supersede · preference · "prefers dark mode" ↻ update
reject · context · low importance (0.18) ✗ drop
pipeline_timing · extract 3487ms · dedup 51ms · persist 124ms
$ curl -X POST https://api.memgram.io/memory/search \
-d '{query:"technical preferences", user_id:"usr_4f1", limit:5}'
2 results · 84ms · hybrid graph + vector
09 · Teams in production

What engineers say after wiring it in.

“Before memgram, debugging long-running conversations felt mostly probabilistic. We could see prompts and traces, but not how the system's memory state evolved over time. The retrieval traces changed that immediately.”
Aria Vance
Core AI Team
“We were chasing a hallucination for weeks. Turned out to be cross-agent contamination, a memory written by support surfacing in the sales agent. memgram showed us the exact transition.”
Daniel Okafor
AI Product Engineer
“We didn't need another vector database. We needed visibility into what our agents actually believed about users across sessions. memgram gave us a much clearer operational model for long-term memory.”
Maya Bergstrom
Staff AI Engineer
10 · Pricing

Simple plans. Scale when your agents do.

Every plan includes the full memgram cognition stack — observability, retrieval traces, and graph memory. You only pay for the writes your agents persist.

Free
$0/month
10,000 memory adds/month
  • Unlimited searches
  • Unlimited agents
  • Unlimited users
  • Hybrid BM25 + vector search
  • Graph memory
  • Full observability
Start free
Starter
$29/month
50,000 memory adds/month
  • Unlimited searches
  • Unlimited agents
  • Unlimited users
  • Hybrid BM25 + vector search
  • Graph memory
  • Full observability
Get started
Most popular
Growth
$129/month
225,000 memory adds/month
  • Unlimited searches
  • Unlimited agents
  • Unlimited users
  • Hybrid BM25 + vector search
  • Graph memory
  • Full observability
Get started
Pro
$299/month
500,000 memory adds/month
  • Unlimited searches
  • Unlimited agents
  • Unlimited users
  • Hybrid BM25 + vector search
  • Graph memory
  • Full observability
Get started
Enterprise
Let's talk.

Unlimited scale, dedicated infrastructure, and policies tailored to your compliance posture. For teams running mission-critical agents in production.

Contact us
  • Unlimited memory adds
  • Unlimited searches, agents & users
  • Dedicated VPC / self-hosted
  • Advanced auth & audit logs
  • Custom policies & SLAs
  • Priority support & onboarding

Stop guessing what your AI remembers.

memgram is the control plane for AI cognition — inspectable state, traceable retrieval, governable persistence. Built for production AI systems.

Start Free Try Playground