Lians CONSOLE
Set up your Lians workspace and give your agents an auditable source of truth.
Sign in securely with Google or GitHub to create your workspace.
Lians CONSOLE
Please wait while we complete your sign-in.
STEP 1 OF 7
Choose the environment you are setting up.
STEP 2 OF 7
STEP 3 OF 7
STEP 4 OF 7
STEP 5 OF 7
STEP 6 OF 7
Optional, but it helps us tailor your workspace.
0 / 20,000STEP 7 OF 7
Everything looks good. You can change these later in Settings.
GET STARTED
Start with local, zero-credential agent memory. Move to a self-hosted environment when you are ready for your own PostgreSQL, KMS, and policy controls.
Skip the manual setup: paste this prompt into Claude Code or Codex and your agent wires Lians into your codebase.
Integrate Lians (regulated AI memory: bitemporal recall, tamper-evident audit, crypto-shred erasure) into this project.
1. Detect the stack and install the SDK:
- Python: pip install lians-sdk (extras: [langchain] [langgraph] [crewai] [mcp] [all])
- TypeScript / Node: npm install @lians-ai/lians
- Go: go get github.com/Lians-ai/Lians/agentmem/sdk/go@v0.4.1
- Java (JVM 11+): ai.lians:lians-sdk:0.4.1 (Maven Central)
2. Configure via environment only (never hardcode secrets): read LIANS_URL and LIANS_API_KEY.
If they are unset, add both to .env.example and tell me to create a key in the Lians console
(API keys view), or self-host: git clone https://github.com/Lians-ai/Lians && docker compose up --build.
3. Wire memory into the agent loop (no agent loop in this project? skip to 4 and integrate at dev time instead):
- Write durable facts after meaningful exchanges: mem.add(agent_id, content, event_time, metadata).
- Recall context before responding: mem.recall(agent_id, query), or mem.recall_at(query, as_of=<ISO date>)
whenever the answer must reflect what was known or valid at a past point in time.
- Put structured keys in metadata (ticker / subject / metric) so supersession can close stale
versions automatically instead of letting revisions pile up.
4. If this project is an MCP host, or you are working inside one (Claude Desktop / Claude Code / Cursor /
VS Code), skip SDK plumbing and register the MCP server: command "uvx", args
["--from","lians-sdk[mcp]","lians-mcp"], env: LIANS_URL, LIANS_API_KEY, LIANS_AGENT_ID.
uvx ships with uv - if missing, install it first: https://docs.astral.sh/uv (winget/brew/curl).
5. Never delete or overwrite memory rows directly: supersession closes stale facts, and
POST /v1/erase (crypto-shred) is the only compliant erasure path.
6. Finish with a smoke test: add a fact with yesterday's event_time, assert recall() returns it, and
assert recall_at(as_of=<a date before the write>) returns nothing. No key or server yet? Run it
serverless: pip install lians-sdk[local], then LocalLiansClient().
Docs: https://www.lians.ai/docs · SDKs: https://www.lians.ai/sdks
PLAYGROUND
Writes and recalls run against your own namespace on the live Lians engine: supersession, REFINES narrowing, admission control, and relevance scores all apply exactly as in production.
NVDA FY2026 revenue guidance revised to $32B on February 20, 2025. Superseded by the May update.
✓ Validity window verified · ✓ Audit event loggedDEVELOPER ACCESS
Create keys for your agents. A secret is shown once, while its name, prefix, creation time, and status remain available here.
Copy this secret now. It will not be shown again.
Create a key for a local agent, a staging environment, or your production service.
Run the Lians local server to create and manage keys.
ACTIVITY
Follow the setup guide, then return here to inspect memory activity.
ACTIVITY
Every memory retrieval will be recorded here with its time filter and provenance.
ACTIVITY
API writes and recalls will appear here when your agent begins using Lians.
ACTIVITY
Entity metadata from your memory records will become inspectable here.
ACTIVITY
When agents write facts, their current validity and source will appear here.
ACTIVITY
Each add, recall, supersession, and erasure event will contribute to the audit chain.
MEMORY GOVERNOR
Memory changes land as governed proposals, not silent overwrites. Low-confidence supersessions (including REFINES narrowings) and writes held by admission control queue here for a human decision before they shape the durable record.
Checking your namespace for flagged supersessions.
Checking for writes held by the admission policy gate.
ACTIVITY
Explore related entities and the relationships between memory updates as your agent writes data.
ACTIVITY
Send signed events to your systems when memories are added, superseded, or erased.
ACTIVITY
Create controlled exports of memory and audit metadata for internal review.
ACCOUNT
ACCOUNT
Loading your plan…
CHOOSE A PLAN
The complete engine is open source. Managed plans price hosted capacity, automation, operations, and support, not permission to use the product’s core capabilities.
MAXIMIZE YOUR USAGE