Remember
Save project facts, decisions, notes and source material in a corpus. Give a memory a stable document ID so you can come back to it.
Memory beyond the conversation
Limitless memory for humans and AI agents
Give your agents memory that lasts beyond a conversation. Store what matters, recall it by meaning, and bring useful context into the next task—with one API and a CLI built for agents.
Managed early access by arrangement. Start with the docs to choose your setup.
The CLI is your agent’s starting point
curl -fsSL https://install.enscrive.io/install.sh | sh Published CLI: Linux x86_64. Installation and prerequisites.
A familiar way to work 01 02 03 04
Remember what matters. Pick up where you left off.
Build a more human-like memory experience: bring prior facts, preferences and decisions into the next task. You control what the agent saves and when it recalls it.
Recall
Ask in natural language. Neural search brings back relevant passages, even when the question uses different words.
Update
Use explicit replacement to revise a document under the same ID. Keep the context your agent retrieves aligned with the work as it changes.
Evaluate
Check retrieval against questions and expected answers. Compare configurations before making them part of your agent’s routine.
Built for agents. Useful to people.
A memory tool your agent can work with.
Use the CLI from a shell or an agent’s tool runner. Structured JSON responses and job status make the workflow scriptable; a developer portal gives people a place to inspect corpora, usage and wallet activity.
The CLI connects to an Enscrive deployment. Install the client, follow the setup guide, and configure the endpoint and credentials for your environment. Self-managed evaluation also needs the services and any selected provider access.
Free self-managed use covers non-commercial exploration, evaluation and development. Production or commercial use requires a subscription or commercial license. Read the license summary.
Read the quickstartRemember, then recall
# After configuring your endpoint and API key
# Use an existing corpus with an explicit embedding model
enscrive ingest documents --corpus-id "$CORPUS_ID" \
--document-id launch-notes \
--content 'The launch review is on Thursday.' --mode replace
enscrive search --corpus "$CORPUS_ID" \
--query 'When is the launch review?' --limit 5 --output jsonUse the current CLI. Choose a corpus and embedding model before ingestion. Ingest waits for its asynchronous job by default.
Corpora + Voices
Memory shaped for your work.
A corpus holds your material and declares its embedding model. A Voice configures how content is chunked and retrieved. Together they let you organize what an agent knows and tune how it finds context.
Project memory
Keep decisions, constraints and working notes close to the next implementation task.
Research memory
Collect sources and retrieve passages as new questions emerge.
Customer context
Bring approved preferences and previous interactions into a support workflow.
Evaluation belongs in the workflow
Know what your agent can find.
Start with a supported retrieval dataset or your own questions and relevance judgments. Ingest its corpus, run retrieval evaluations, and compare the results as you tune the configuration.
Measure the memory you need
Recall, precision, nDCG and MRR help you inspect retrieval quality. Use representative questions and review the returned sources, especially when the next action depends on them.
Scores apply to the dataset and configuration tested; evaluate questions your agents will actually face.
Explore the documentationOne public API
Put memory inside your application.
Use HTTP and JSON to ingest documents, search a corpus, manage configurations and inspect jobs. Keep your agent’s reasoning in your application; let Enscrive supply the context.
Managed early access is available by arrangement. Connect with the endpoint and API key supplied for your deployment.
Authentication and API docsRecall through HTTP
# Use the endpoint and API key supplied for your deployment
curl "$ENSCRIVE_BASE_URL/v1/search" \
-H "X-API-Key: $ENSCRIVE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"corpus_id":"YOUR_CORPUS_ID", "query":"When is the launch review?", "limit":5}'The toolkit
From saved facts to useful context.
Memory with a home
Organize knowledge into corpora for a project, a purpose or a body of source material.
Search by meaning
Retrieve useful context for the next task without pasting an entire archive into a prompt.
Choose your model
Select an embedding model for each corpus. Provider choice and processing requirements stay explicit.
Shape the context
Use Voices to configure chunking and retrieval behavior for the material your agent works with.
Keep sources close
Retain document identity and metadata so an agent can connect a retrieved passage to its source.
Measure retrieval
Use datasets and relevance judgments to compare recall and ranking for your own workload.
Work from the terminal
An agent can discover commands, ingest content, search and inspect jobs through the CLI.
Inspect usage
Review usage records and exact wallet amounts alongside readable display values.
Build on one API
Connect your application to the public /v1 surface, with asynchronous jobs for ingestion.
Before you begin
A clear path into early access.
What does “limitless” mean?
It is our vision: useful memory beyond a single conversation or context window. Storage, rate limits and deployment capacity still apply. Your application chooses what to save; retrieval is not perfect recall or autonomous learning.
How do I get access?
Contact us for managed early access and workload fit. For self-managed evaluation, follow the installation and setup guides. Published plan prices are on the pricing page; availability and commitments need confirmation.
Where does my content go?
Selected embedding and language-model providers receive content needed to process a request. A deployment’s hosting region does not guarantee that all provider processing stays there. Review privacy and security before choosing your setup.
How can I inspect cost and changes?
Usage records and the wallet expose activity and exact pico-dollar amounts. A small debit may display as $0.00 alongside its exact value. Replacement updates retrieval context; removal of old storage and historical files is a separate lifecycle. See billing accuracy.
Bring the past into the next task