Dont Observe - Audit your AI Agent.
Operational database for AI agents — replay sessions, trace decisions, detect drift.
Open Source ↓ · Pro ↓ · Team ↓ · Docs · Live site
↑ Activity · Behavior · Reports · Suggestions — same dashboard on managed cloud and after quickstart
| Open Source | Pro | Team | |
|---|---|---|---|
| Best for | Self-hosters & contributors | Solo devs & early prod | Teams with multiple agents |
| Price | Free forever (AGPL) | €29 / month | €69 / month |
| Hosting | Your machine / VPC | db.zizka.ai | db.zizka.ai |
| Events / month | Unlimited (your infra) | 50k | 100k |
| Projects | Unlimited | 2 | 5 |
| Dashboard | Activity · Behavior · Reports · Suggestions | Same + Fleet (managed) | Same + Fleet + priority support |
| Support | Community | Priority | |
| Jump to | 4 steps ↓ | 4 steps ↓ | 4 steps ↓ |
Enterprise VPC? Single-tenant deploy, commercial license → db.zizka.ai/enterprise
This repo is the open-source product runtime (API, dashboard, SDKs, MCP). Managed operator tools live in a private repo; all product links point here.
Free · AGPL-3.0 · Your infrastructure
| Step | What you get |
|---|---|
| ① What it is | ZizkaDB stores every agent step as a linked event — not scattered logs. Trace decisions with why(), rewind with at(), search history in plain English, and catch behavioral drift before users notice. You run the full stack: API + dashboard + SDKs. |
| ② How to integrate | Run locally in ~2 min: curl -fsSL …/quickstart-remote.sh | bash then pip install zizkadb-sdk (or TS / MCP / REST). Link events with parent_id so why() can walk the chain. → Full connect guide |
| ③ Documentation | CONNECT.md · Docs · Self-hosting · Examples · Architecture |
| ④ Live dashboard | localhost:3001/login → Open my dashboard (no signup). Log from SDK → refresh → see sessions live. |
Quickstart
curl -fsSL https://raw.githubusercontent.com/Zizka-ai/ZizkaDB/main/scripts/quickstart-remote.sh | bashasync with ZizkaDB(host="http://localhost:8000") as db:
user = await db.log(agent="my-bot", event="user_message", data={"text": "Why is my order late?"})
tool = await db.log(agent="my-bot", event="tool_call", data={"tool": "lookup_order"}, parent_id=user.event_id)
await db.why(tool.event_id).print()
Causal chain replay with db.why() after quickstart
Hosted for you · No Docker to maintain
| Step | What you get |
|---|---|
| ① What it is | Same engine, hosted on db.zizka.ai. Replay sessions and trace causal chains without running your own databases. vs OSS: zero infra · instant API keys · Fleet tab on managed cloud. |
| ② How to integrate | Sign up → · create agent · copy API key · pip install zizkadb-sdk · log with api_key="zizkadb_live_...". Also: npm, MCP, REST. → CONNECT.md |
| ③ Documentation | db.zizka.ai/docs · Swagger · Community · 50k events/mo · 2 projects · email support |
| ④ Live dashboard | db.zizka.ai/dashboard — sign up, log from SDK, sessions appear in seconds. Same UI as self-host, always online. |
async with ZizkaDB(api_key="zizkadb_live_...") as db:
await db.log(agent="support-bot", event="user_message", data={"text": "Hello"})
Multiple agents & projects in production
| Step | What you get |
|---|---|
| ① What it is | Everything in Pro, with room to grow. vs Pro: 100k events (vs 50k) · 5 projects (vs 2) · priority support · built for multi-agent ops with Fleet ranking. |
| ② How to integrate | Sign up for Team → · one agent per service · consistent session_id per conversation. → Multi-agent wiki |
| ③ Documentation | Plan picker · Docs · Production wiki · 100k events/mo · 5 projects · priority support |
| ④ Live dashboard | db.zizka.ai/dashboard — switch agents, compare baselines, Fleet ranking across your workspace. |
const db = new ZizkaDB({ apiKey: process.env.ZIZKADB_API_KEY! })
await db.log({ agent: 'sales-bot', event: 'tool_call', data: { tool: 'crm_lookup' }, parentId: prev.eventId })Core unit tests plus the SDK and MCP package tests do not require a running ZizkaDB stack:
# Core tests
(cd core && python -m pip install -r requirements.txt -r requirements-dev.txt && python -m pytest tests -q)
# Python SDK tests
(cd sdk/python && python -m pip install -e '.[dev]' && python -m pytest tests -q)
# MCP tests
(cd mcp && python -m pip install -e '.[dev]' && python -m pytest tests -q)
# TypeScript SDK tests
(cd sdk/typescript && npm ci && npm test)| Capability | OSS | Pro | Team |
|---|---|---|---|
db.log() · db.why() · db.at() |
✓ | ✓ | ✓ |
| Semantic search & drift baselines | ✓ | ✓ | ✓ |
| Activity · Behavior · Reports · Suggestions | ✓ | ✓ | ✓ |
| Fleet tab (managed cloud) | — | ✓ | ✓ |
| Self-host / AGPL | ✓ | — | — |
| Managed hosting | — | ✓ | ✓ |
| 50k events / month | — | ✓ | — |
| 100k events / month | — | — | ✓ |
| 2 projects | — | ✓ | — |
| 5 projects | — | — | ✓ |
| Priority support | — | — | ✓ |
| Enterprise VPC | Contact → |
| Python | TypeScript | LangChain | CrewAI | MCP (Cursor) | REST |
|---|---|---|---|---|---|
pip install zizkadb-sdk |
npm i zizkadb-sdk |
Guide → | Guide → | uvx zizkadb-mcp |
Swagger → |
All SDK examples (Python · TypeScript · LangChain · CrewAI · MCP · REST)
See CONNECT.md for full copy-paste examples.
pip install zizkadb-sdk
npm install zizkadb-sdk
uvx zizkadb-mcp
zizkadb init my-agent --template basic| Function | What it does |
|---|---|
db.log() |
Record any agent step; use parent_id for causal links |
db.why() |
Walk backward from any event to root cause |
db.search() |
Semantic search over agent history |
db.at() |
Reconstruct state at a point in time |
db.baseline() |
Detect behavioral drift |
db.context_for() |
Inject relevant past events into prompts |
db.forget() |
GDPR erasure by metadata filter |
| Resource | Link |
|---|---|
| Issues | GitHub Issues |
| Discussions | GitHub Discussions |
| Forum | db.zizka.ai/community |
| Contributing | CONTRIBUTING.md |
| Security | SECURITY.md |
| License | API & dashboard AGPL-3.0 · MCP MIT |
 |
Disable telemetry: export ZIZKADB_TELEMETRY=false