YantrikDBCAUTION
Cognitive memory database for AI agents — consolidates duplicates, detects contradictions, fades stale memories via temporal decay. Rust, Apache-2.0, ships as library / MCP server / HTTP cluster.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
⚠ Correction notice (2026-04-19): Phase 3 benchmark writeups committed earlier today (docs/phase3a/,docs/phase3b/,docs/phase3c/,docs/phase3d/) used a Python simulator for the "structured memory" condition — NOT the actual yantrikdb engine. Rerun with real yantrikdb is indocs/phase3e/. Full audit trail, preliminary corrected numbers, and the exact publication gate for a findings post are inCORRECTIONS.md.
A memory database that forgets, consolidates, and detects contradictions.
Vector databases store memories. They don't manage them. After 10,000 memories, recall quality degrades because there's no consolidation, no forgetting, no conflict resolution. Your AI agent just gets noisier.
YantrikDB is different. It's a cognitive memory engine — embed it, run it as a server, or connect via MCP. It thinks about what it stores.
Shortest path to try it: MCP setup for Claude Code / Cursor / Windsurf → (one pip install, one config block).
The bigger picture: YantrikDB is the memory layer being built on the road to YantrikOS — an AI-native operating system where agents are first-class primitives, not apps on top. Memory was the bottleneck, so we're shipping it first.
[](https://crates.io/crates/yantrikdb-server) [](https://pypi.org/project/yantrikdb/) [](https://github.com/yantrikos/yantrikdb-server/pkgs/container/yantrikdb) [](https://doi.org/10.5281/zenodo.20128887) [](LICENSE)
📄 New paper (May 2026): [*Skill as Memory, Not Docume
f97bf6065639OBSERVED · 2026-10-06Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add yantrikdb --env YANTRIKDB_API_KEY=${YANTRIKDB_API_KEY} --env YDB_TOKEN=${YDB_TOKEN} -- uvx yantrikdb{
"mcpServers": {
"yantrikdb": {
"command": "uvx",
"args": [
"yantrikdb"
],
"env": {
"YANTRIKDB_API_KEY": "${YANTRIKDB_API_KEY}",
"YDB_TOKEN": "${YDB_TOKEN}"
}
}
}
}Exposed tools (14)
11 read · 3 write · 0 destructive.
| Tool | Risk | Description |
|---|---|---|
conflict_list | read | List memory conflicts (contradictions) that need resolution. |
conflict_resolve | read | Resolve a memory conflict using a strategy. |
entity_edges | read | Get all relationships for an entity in the knowledge graph. |
entity_relate | write | Create a relationship between two entities in the knowledge graph. |
memory_correct | read | Correct an existing memory with updated information. |
memory_forget | read | Permanently forget (tombstone) a memory. |
memory_get | read | Get a specific memory by its ID. |
memory_recall | read | Search memories by semantic similarity to a natural language query. |
memory_recall_feedback | read | Provide feedback on a recall result for adaptive learning. |
memory_recall_refine | read | Refine a previous recall with a follow-up query. |
memory_record | read | Store a new memory in the cognitive memory engine. |
memory_stats | read | Get current memory engine statistics. |
memory_think | write | Run the cognitive maintenance loop on the memory store. |
trigger_list | write | Get pending proactive triggers from the memory system. |
Trust audit
CAUTIONgrade C · trust 75/100 Install with care. The audit found things worth knowing before you trust its output.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | WARN |
| L1 | Static analysis of the code | WARN |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (7 observation(s))
- Network
- none-observed
- Shell
- none-observed
- Dependencies
- pinned
- Secrets in source
- found
Findings (24)
melfilters.bytes
let token = "ydb_test_token_for_integration_tests_only_1234567890abcdef";
cluster-secret: "change-me-to-a-random-string"
TOKEN = "ydb_31416c8f210b15b679172740795a295cef75b796aeb521dc3d018d98ea9fbadd"
TOKEN = "ydb_1bf25b1e8e301a7b1906812a0fe62126967e69f51ae2ae3fe25cef4dfda8f4f3"
.put_content("tenants/1/../../etc/passwd", &[0])assert!(store.path("../../etc/passwd").is_none());assert!(!store.has("../../etc/passwd"));curl -X POST http://192.168.4.180:7438/v1/remember \
curl -X POST http://192.168.4.180:7438/v1/recall \
db = connect("http://192.168.4.180:7438", token="ydb_cluster_3a8f...")yantrikdb cluster status --url http://192.168.4.180:7438 --token <secret>
URL=http://127.0.0.1:7438
hlc_bytes = bytes.fromhex(since_hlc) if since_hlc else None
op["hlc"] = bytes.fromhex(op["hlc"])
op["embedding_hash"] = bytes.fromhex(op["embedding_hash"])
peer_actor, bytes.fromhex(hlc_hex), op_id
op["hlc"] = bytes.fromhex(op["hlc"])
I characterize these failure modes formally and measure all three on a 5,000-skill corpus. I report a database-native resolution implemented as YantrikDB v0.8.13, an open-source skill substrate that s
I characterize these failure modes formally and measure all three on a 5,000-skill corpus. I report a database-native resolution implemented as YantrikDB v0.8.13, an open-source skill substrate that s
crates/yantrikdb-wasm/pkg/yantrikdb_wasm_bg.wasm
src/yantrikdb/_yantrikdb_rust.cp313-win_amd64.pyd
uv.lock
Phase 1: `ydb_<random_64_hex>` → maps to one database, full access.
Gates applied: no_behavioural_pass.
f97bf6065639full audit observations/trust-audit/mcp-server/yantrikos__yantrikdb-1.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-06 | f97bf6065639 | CAUTION | C | 75 | first audit |
Questions
What is the YantrikDB MCP server?
Cognitive memory database for AI agents — consolidates duplicates, detects contradictions, fades stale memories via temporal decay. Rust, Apache-2.0, ships as library / MCP server / HTTP cluster.
What tools does YantrikDB expose?
14 in total: 11 read-only, 3 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is YantrikDB safe to connect to an agent?
With care. The audit graded it C (75/100) and found 24 things worth knowing before you trust this server, listed below with the exact line each was found on.
What credentials does YantrikDB need?
It reads YANTRIKDB_API_KEY and YDB_TOKEN from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does YantrikDB run?
It speaks stdio, so it runs as a local process your client starts. It is published on PyPI as yantrikdb.
How current is this page?
The grade is for one exact copy of the source (f97bf6065639), read on 2026-10-06. The repository is watched and re-audited when it changes.