JimoTablelite
中文Run Free Edition
⚡ 100% Local-First & Postgres Native

Fastest data runtimefor AI agents with human-in-the-loop UI

JimoTable is the Postgres-backed runtime built for sharing structured relational data between your AI agents and human teammates, without hand-rolling frontend CRUD panels. Zero cost, zero config. Ship agent workflows in minutes instead of weeks.

Run with Docker(yes, free)Try Live Demo Sandbox
$git clone https://github.com/jimotable/jimotable.git && cd jimotable && ./quickstart.sh
jimotable-runtime :: localhost:8080
LIVE AGENT CONNECTED
agent-session — bashPG16
🤖 [Agent]Received inbound RFP payload (14.2 KB)
🔍 [Agent]Deduplicating customer: 'Ant Tech'... Found (cust_98827301)
✍️ [Agent]Inserting follow-up activity with 1:N relation link... Done (8ms)
⚡ [Agent]Dispatched to sales rep: Sarah Chen (view: Rep_Sarah_Pending)
Schema: base_bse_crmQPS: ∞
Customers (Master)
Auto-Synced
Company
Industry
Rep
Status
Ant Technology
FinTech & AI
Sarah Chen
Follow-up Added
BytePulse AI
Enterprise SaaS
Marcus Vance
Proposal Sent
NovaCloud Inc.
Cloud Infrastructure
Elena Rostova
Negotiation
Ant Technologycust_98827301
Detail Drawer
Linked Activities (1:N)2 events
🤖 Agent WorkerToday 14:32

Agent auto-parsed RFP inquiry and summarized 4 core requirements.

Sarah ChenYesterday 11:00

Quarterly review call with CTO completed.

Linked Deals (1:N)$48,000 / yr
2026 Enterprise API LicenseIn Legal Review
Slide 1•Agent Relational Blueprints

What you can do with JimoTable (lite)

As an agent data runtime, it lets your agents store, query, and link complex business records — and lets your human team audit, edit, and dispatch in one place.

📇Sales & RevOps

AI Sales CRM & Pipeline

Autonomous customer deduplication, agent meeting log synthesis, and automated sales rep assignment into tailored review views.

Autonomous Agent Prompt
"Check if Acme Corp exists, append today's discovery call summary, and assign to David."
📡Market Intelligence

Bidding & Feeds Monitor

Multi-source tender scrapers writing thousands of competitive updates every minute directly to Postgres without rate limit crashes.

Autonomous Agent Prompt
"Scrape provincial tender board, extract budget and qualification criteria, link to Source #4."
✍️Content Matrix

Viral Content & Multi-Platform

From a single topic outline, let agents fan out into platform-tailored copywriting, generate AI banners, and stage posts for approval.

Autonomous Agent Prompt
"Draft 3 tweet variations and 1 newsletter draft from the release notes topic, attach preview PNGs."
🎫Support & Ops

Ticket Triage & Escalation

Intent recognition on incoming customer tickets, automated category tagging, and smooth human agent takeover via the side drawer.

Autonomous Agent Prompt
"Classify incoming ticket #4092, diagnose error logs, and flag for Tier-2 engineering review."
🔍Procurement

Supply Chain & Leads Discovery

Bulk company registration cleaning, multi-tier supplier relationship mapping, and dynamic tagging for purchasing managers.

Autonomous Agent Prompt
"Extract semiconductor supplier relationships from patent registry and populate 1:N vendor links."
🚀Extensible

Custom Agent Systems

Zero-friction schema creation and automated OpenAPI specs for any ad-hoc internal tool your team envisions.

Autonomous Agent Prompt
"Create base 'inventory_sync' with tables Warehouses and SKUs, return management URL."
Slide 2•Zero-Driver Agent Parasitism

Any agent, connect via jimotable skill

jimotable is the skill that turns your agent into an autonomous data worker. Install it once, and every agent in your workflow drives it directly: Claude Code, Cursor, Dify, Codex...

Install as a global agent skillSkill Registry
$npx skills add jimotable/workspace

Compatible with any framework adhering to the Agent Skills Protocol or OpenAPI 3.0 tool definitions.

Works out of the box with
Claude Code
Native CLI Tool integration
Cursor
Agent rules & project skills
Dify / FastGPT
Custom OpenAPI tool nodes
Custom Python / Node
Direct REST / Token calls
Natural Language Agent Orchestration
Autonomous Execution
👤 User
"Build a CRM for electric vehicle manufacturers with Customers and VisitLogs tables, and record the 3 EV makers we just analyzed."
🤖 Agent Tool Calling (jimotable-workspace)
1.Calling jimotable-workspace.create_base(name='ev_crm')
2.Creating table 'Customers' with fields [name, country, battery_tech]
3.Creating table 'VisitLogs' with 1:N relational link to 'Customers'
4.Inserting 3 companies (BYD, Nio, Rivian) and generated discovery logs
✅ Agent Output
✓ Done! Ready for your team at http://localhost:8080/base/bse_ev_crm
Slide 3•High Throughput

Parallel multitasking, execute faster.

Cloud spreadsheet SaaS tools (Feishu, Airtable) enforce harsh artificial rate limits (5–20 QPS) that cause batch-crawling agents to crash with 429 Too Many Requests. JimoTable runs on local PostgreSQL 16 with millisecond commits and zero artificial bottlenecks.

0 QPS Walls
Pure local PostgreSQL throughput
Concurrency & Rate Limits Benchmark
❌Cloud SaaS (Feishu / Airtable)HTTP 429 Error

Feishu / Airtable: 5~20 QPS limit → 429 Rate Limit Crash

⚡JimoTable Free Runtime0ms Queuing

JimoTable: Native PG engine → Millisecond atomic commits

PostgreSQL Physical Storage Anatomy
database: jimotable_cluster
📁 schema: base_bse_crm(Isolated DDL)
├── table: tbl_customer [B-Tree Primary Key, Foreign Constraints]
├── table: tbl_activities [1:N Indexed FK -> tbl_customer]
└── table: tbl_orders [Atomic ACID Transaction]
VS Low-Code Bloat: Single generic table `public.records` with `data JSONB` column — zero physical indexes, heavy GC vacuum bloat, global lock on schema changes.
Slide 4•Storage Engine

Real relational tables, not giant JSONB.

Many low-code tools dump dynamic rows into a monolithic JSONB column in one giant table, killing performance, indexing, and transactional integrity. JimoTable uses a Base-per-schema architecture: each Base gets an isolated PostgreSQL schema with native B-Tree indexes and foreign keys.

100% Native
Isolated Postgres schema per Base
Slide 5•Human-in-the-loop

Drawer for humans, API for agents.

Flat spreadsheets fail to show the complete picture of modern relational business. JimoTable features an instant Relational Detail Drawer: click any record to inspect all linked subrecords, activity timelines, and audits. Edit or dispatch right inside the drawer.

0 Code UI
Pre-built spreadsheet & relational drawer
Entity Detail Drawer (Zero-Context Switching)1:N Slideout
Human-in-the-loop Workflow
Clicking any row in the parent table doesn't take the user to another page. The right drawer slides open seamlessly with all related records, communication history, and real-time dispatch options.
⚡ For AI Agents
Fast append, bulk mutate, deduplication queries via OpenAPI.
👥 For Humans
Inspect timeline, edit cell, approve proposal, dispatch task.
Slide 6•Architectural Matrix

Stop bouncing between incomplete tools.

JimoTable replaces hand-rolled Vue CRUD dashboards, SaaS tables with QPS walls, and bare ORM glue.

Feature / Dimension
⚡JimoTable (Free)
Feishu BitableNocoDBSupabaseHand-Rolled CRUD
Built-in Human Admin UI
✓Native Excel Grid + Drawers
✓Built-in SaaS UI
✓Basic Grid UI
—DB dashboard only (no business UI)
—Requires 2 weeks frontend coding
Relational Detail Drawer (1:N)
✓Instant slideout linked drawer
—Popup modal only
—Sub-table list popup
—None (build custom UI)
—Manual Drawer component
No QPS Rate Limit (Agent Friendly)
✓Zero artificial limits (PG speed)
—Strict limit (5~20 QPS, 429 error)
✓Local throughput
✓Postgres throughput
✓Depends on server capacity
100% Local-First & Private VPC
✓Pure Docker / Private VPC
—Cloud SaaS (data in vendor cloud)
✓Self-hosted
✓Self-hostable
✓Self-hosted
Native Agent OpenAPI Contract
✓Token-optimized OpenAPI & Skills
—Generic heavy REST API
—Standard REST API
—PostgREST API
—Manual controller endpoints
Autonomous Agent Controllable
✓Direct npx skills add jimotable
—Complex OAuth / App credentials
—API token only
—Requires SQL / ORM wrapper
—Manual prompt engineering
Community Free & Self-Hosted
✓Free Community Edition
—Per-seat SaaS subscription
✓Open Source
✓Open Source
—High dev & maintenance cost
Quickstart•60 Seconds Launch

⚡ 60-Second Quick Start

Launch your local agent runtime with zero configuration.

1

1. Launch with One Command

Run the zero-config script to pull official images and generate secure tokens automatically:

git clone https://github.com/jimotable/jimotable.git
cd jimotable
./quickstart.sh
2

2. Open Local Web Workspace

Visit http://localhost:8080 in your browser to inspect the clean, high-density spreadsheet interface.

http://localhost:8080Open ↗
3

3. Run Agent Demo Simulation

Watch an autonomous agent deduplicate records and link relational logs in real-time:

# Autonomous Agent Simulator
./demo/simulate-agent.sh
Slide 7•FAQ

Frequently Asked Questions

Everything you need to know about JimoTable and agent workflows.

What is JimoTable (lite)?↓
JimoTable (lite) is a private, Postgres-backed business data runtime built specifically for AI agents. It gives agents high-throughput OpenAPI pipelines to store and link records, while simultaneously providing human business teams with an Excel-like management UI with instant relational drawers — saving two weeks of hand-rolling frontend CRUD dashboards.
How does it compare to Feishu Bitable or Airtable?↓
Feishu and Airtable are public cloud SaaS applications designed for manual human input. They impose severe QPS rate limits (e.g. 5–20 QPS) that break batch-processing agents with HTTP 429 errors, and your private business data is locked in their cloud. JimoTable runs 100% locally on your own Postgres 16 instance with zero artificial rate limits and strict privacy compliance.
How does it compare to Supabase or bare PostgreSQL?↓
Supabase provides excellent backend database APIs, but it doesn't give your non-technical teammates an out-of-the-box management UI. You still have to spend days or weeks building Vue/React admin panels, table views, and dispatch drawers. JimoTable includes the full human-in-the-loop web interface right away.
How does it compare to NocoDB or Baserow?↓
NocoDB is a no-code database tool built primarily for human operators. JimoTable is designed from day one for AI Agents: token-optimized schema contracts, asynchronous non-blocking DDL migrations, full-trace Tool Call auditing, and 1:N relational slideout drawers tailored for agent-human handoffs.
Can JimoTable work with Claude Code, Cursor, and Dify?↓
Yes! Any agent or LLM workflow capable of executing shell commands or calling HTTP REST endpoints can drive JimoTable. You can attach our pre-packaged `jimotable/workspace` skill in one command: `npx skills add jimotable/workspace`.
Is my business data secure?↓
100% secure. JimoTable runs entirely in your local Docker environment or private company VPC. It never sends your business data, customer records, or agent activity logs to external cloud servers.