Live context for agents

Coalent turns your sources into dense, attributed facts — served at a fraction of RAG’s tokens and invalidated the instant a source changes.

Never stale.Always attributed.

pricing.md$29 $39fresh 0.4s
pip install coalent
Mount in Claude Code →

// apache-2.0 · zero-dependency core · built in the open

[01]PROOFmeasured, not projected
ISO-TOKEN0.731 @ 981 toknaive’s best accuracy at 43% fewer tokens
COALENT
981
NAIVE k12
1,729
RANKING1 / 6 / 15gold fact ranks (p50 / p75 / p90)
RANK 1–20
p50p75p90
BEHAVIOR−33%refusals with the behavioral stack
BASELINE
100
W/ STACK
67
REUSE0 61%repeat asks: first-pass answers
FIRST ASK
0%
REPEAT
61%

n=605 held-out · strict grading · third-party gold questions — method →

[02]WHERE COALENT SITSthe honest map
YOUR AGENT
HeadroomTRANSPORT

compresses the bytes of each call — per-request, stateless.

ACEPROCEDURAL

evolves an agent’s playbook — how to act, not what’s true.

Mem0USER MEMORY

remembers the person across sessions — who you’re talking to.

CogneeCORPUS GRAPHS

builds knowledge graphs over your corpus, upfront.

CoalentFRESH SOURCE FACTS

dense claims derived from your sources, invalidated by provenance the instant a source changes.

YOUR SOURCES

Coalent doesn’t compete for who-you-are or how-to-act — it complements the memory layers and owns the freshness slot: what your sources say right now, with context invalidation that is surgical — span-level provenance, not a TTL guess. That’s why its layer sits directly on your sources.

[03]HOW IT WORKSone loop, no upfront build
[01]

Sources

docs · tickets · APIs

Point the cache at wherever your knowledge lives. Units build lazily, on first read — no upfront graph tax.

[02]

Atomic claims

extract once · span-attributed

Each source becomes query-independent claims — every number kept, every claim pinned to the span it came from.

[03]

Budget-packed serving

read_path="pool"

Every read gets the globally-ranked fresh claims, packed to your token budget under per-source attribution headers.

[04]

Surgical invalidation

source Δ → its claims only

A changed source stales exactly the claims derived from it. They rebuild on the next read — nothing else is touched.

[04]INTEGRATIONSBYO everything

The same cache behind an MCP context server for Claude Code and Cursor, or as a LangChain cache over the vector store, embeddings, and chat model you already run. Nothing about how you built them changes.

  • CLAUDE CODE
  • CURSOR
  • MCP
  • LANGCHAIN
  • ANY VECTOR DB

MCP — CLAUDE CODE · CURSOR · ANY CLIENT

terminal
$ claude mcp add coalent \
-- coalent-mcp --watch ./docs
The MCP server →

LANGCHAIN — YOUR EXISTING STACK

python
cache = create_coalent_cache(
vectorstore, llm=llm,
embeddings=emb)
langchain-coalent →

// folder mode (--watch) is zero-setup with a measured cost: 0.46 vs 0.71 in factory mode — bring your own retriever for the full number

[05]WHAT THE CACHE HOLDSthree properties, one substrate

Dense.

Atomic claims, budget-packed — same answers, ~half the context.

Durable.

Persists across sessions, agents, and users — and learns from refusals.

True.

Span-level provenance; a changed source invalidates exactly its claims.
source Δits claims stalere-extractfresh
[06]BENCHMARKparity, not a beat

We measured naive RAG’s own token-scaling curve on the same stream — so Coalent is compared against every operating point of the baseline, not one convenient k. Real news corpus, third-party gold questions, n=605 held-out, strict grading.

ACCURACY

Coalent pool0.731 @ 981 tok

CONTEXT TOKENS / READ

ARMACCURACYCTX TOK
naive k=40.582590
naive k=60.638882
naive k=90.7111,311
naive k=120.7311,729
Coalent pool0.731981

The honest framing: we do not claim to beat naive RAG on accuracy — its own curve reaches the same number given more tokens. The claim is equal accuracy at 25–43% fewer context tokens, plus what no k can buy: stale RAG answers are masked the instant a source changes, and every fact carries its span.

Full methodology →

[07]BUILT IN THE OPENreal surfaces only

v0.6.1 aug 2026 · coalent-mcp + langchain-coalent  ·  v0.6 jul 2026 · read_path="pool"

get started

# dense, durable, true — live context for your agents

pip install coalent