Lore Context
All your AI agents.One shared memory.
Hosted beta memory for Claude Code, Codex, Cursor, and OpenCode — with browser capture clearly labeled beta, plus developer docs, REST API, MCP examples, pricing, and source-backed receipts.
Lore Context
shared memory cloud
bidirectional context capture · source-backed memory return
LongMemEval-S Retrieval
500 questions. Utilizes the same session hit protocol.
Lore context is measured based on the publicly available LongMemEval-S benchmark.
2026-05-04 · local adapternot official QA score
Questions500abstention types excluded
Recall@596.6%session hit in top 5
Recall@1098.8%session hit in top 10
MRR91.5%mean reciprocal rank
NDCG@1090.1%ranking quality
P95 latency23msoffline adapter, not production SLA
01 / Who we are
A memory company for AI agents.
ChatGPTOpenAI
Claude CodeAnthropic
CursorAnysphere
Gemini CLIGoogle
Codex CLIOpenAI
jcode1jehuang
OpenCodeopencode-ai
Clinecline
Roo CodeRooCode
Kilo CodeKilo-Org
GooseBlock
AiderAider-AI
WindsurfCodeium
Lore Contextagent ecosystem
Claude DesktopAnthropic
HermesNous Research
OpenClawopenclaw
02 / What we provide
One service: shared memory for your AI agents.
Temporal graph
Release memory graph
A living graph links raw agent sessions, durable facts, entities, and recall paths with source-backed time windows.
03 / Service quality
Quality means the memory can explain itself.
Lore keeps origin, time, confidence, and control close to every remembered fact.
SourceClaude Code · project session
SavedReviewed preference · not raw transcript
Used byCodex, ChatGPT, Cursor
origin attached
time aware
delete path
Memory receipt
Why this answer?
SourceSaved from a verified working session.
ReasonThe same preference appeared across connected agents.
ControlPause, export, or delete from one vault.
04 / Competitive field
Lore Context vs the field.
Methodology-first comparison using public benchmark notes and each project's own docs. Lore does not claim a production SLA or benchmark win from offline adapter numbers.
System
Benchmark dataset
Recall@5top 5
Recall@10top 10
MRRrank quality
NDCG@10ranking quality
P95 latencyreported path
lore context
LongMemEval-S500 questions
96.6%
98.8%
91.5%
90.1%
23msoffline adapter
agentmemory
LongMemEval-S500 questions, BM25 + Vector
95.2%
98.6%
88.2%
87.9%
not published
MemPalace
LongMemEvalhybrid v4 held-out
98.4%
99.8%
N/A
93.8%
not published
mem0
LongMemEval-SAgentMemory field R@5
81.4%
N/A
N/A
N/A
not published
zep
LongMemEval QA71.2% overall
N/A
N/A
N/A
N/A
2.58sreported, not p95
memobase
LoCoMo75.78% LLM judge
N/A
N/A
N/A
N/A
not published
supermemory
LongMemEval85.2% overall
N/A
N/A
N/A
N/A
<300msp95 claimed
letta
LongMemEval-SAgentMemory field R@5
73.8%
N/A
N/A
N/A
not published
cognee
LongMemEval-SAgentMemory field R@5
78.1%
N/A
N/A
N/A
not published
lore context LongMemEval-S
- Recall
- 96.6% @5, 98.8% @10.
- Rank
- MRR 91.5%, NDCG@10 90.1%.
- P95
- 23ms, offline adapter.
agentmemory BM25 + Vector
- Recall
- 95.2% @5, 98.6% @10.
- Rank
- MRR 88.2%, NDCG@10 87.9%.
- P95
- not published.
MemPalace LongMemEval
- Recall
- 98.4% @5, 99.8% @10.
- Rank
- NDCG@10 93.8%; MRR not published.
- P95
- not published.
mem0 LongMemEval QA
- Recall
- 81.4% @5 from AgentMemory field comparison.
- Rank
- Recall@10, MRR, and NDCG not published there.
- P95
- not published.
zep LongMemEval QA
- Dataset
- 71.2% overall.
- Retrieval
- Recall@K, MRR, and NDCG not published.
- Latency
- 2.58s reported, not p95.
memobase LoCoMo
- Dataset
- 75.78% LLM judge.
- Retrieval
- Recall@K, MRR, NDCG, and p95 not published.
supermemory LongMemEval
- Dataset
- 85.2% overall.
- Retrieval
- Recall@K, MRR, and NDCG not published.
- P95
- <300ms claimed.
letta Letta Evals
- Recall
- 73.8% @5 from AgentMemory field comparison.
- Rank
- Recall@10, MRR, NDCG, and p95 not published there.
cognee HotPotQA
- Recall
- 78.1% @5 from AgentMemory field comparison.
- Rank
- Recall@10, MRR, NDCG, and p95 not published there.
Benchmark methodology: Lore Context / agentmemory = LongMemEval-S retrieval-only recall_any@K; MemPalace = LongMemEval retrieval recall; mem0 / Letta / cognee R@5 = AgentMemory public field comparison on LongMemEval-S; Zep = LongMemEval QA plus reported latency; memobase = LoCoMo LLM judge; supermemory = LongMemEval overall score from public site copy. Offline adapter latency is not a hosted production SLA.
05 / Pricing
Simple, transparent pricing.
Free
Start
$0
Go
Daily API
$9.99/mo
Plus
Heavy recall
$19.99/mo
Pro
Power user
$99.99/mo
Overage is optional.Plans stop at hard caps by default. Turn on paid overage only when you want uninterrupted sync.
$5 / extra 10K add memoriesAdd memories means new memories written to your cloud vault.$5 / extra 10K retrieve memoriesRetrieve memories means searches that return context to an agent.