TabTin
Open-source team harness where humans and multiple AI sub-agents share one task surface for code, docs, spreadsheets and browser research.
If your team has been burned by autonomous agents writing to production without a trail, TabTin's checkpoint-plus-approval model is the right shape. It is genuinely differentiated on the team layer — shared resources, rollback before writes, chat-native task tracking — not on raw model capability. The homepage's claim to be the 'only project to pass' a 28-candidate code-level audit is self-marketing and should be treated as such. Its practical weakness is surface area: the scrape shows no pricing table, no public changelog or docs beyond a developer-support Prompt library, and primary UI copy in Chinese. Pilot it against Cursor or Devin on one real delivery before betting a pipeline on it.
Verified 7d ago · liveness 60/100 · cite: rightaichoice.com/tools/tabtin
- Small product teams that already run on group chat and want agents inside that workflow
- Engineering teams that need auditable, rollback-able agent execution before production writes
- Projects mixing coding, research, documents and spreadsheets on one shared task surface
- Organizations that require human sign-off before any publish or write action
- Solo users wanting a lightweight personal AI assistant
- Teams needing a mature ecosystem of third-party integrations out of the box
- Users who require a documented public API or CLI for custom automation
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip TabTin if you want a hosted, English-first personal AI assistant with a broad third-party integration marketplace — TabTin is a self-hosted, Chinese-language-first team harness whose value depends on your team already collaborating in group chat.
The free tier is a download of the open-source harness plus 48 Skills and agent chat — the Team tier is credits-based, so heavy agent runs move you onto a metered allowance tracked in organization settings.
TabTin prices as freemium with a free open-source download and a credits-based Team tier whose price is quoted as Custom — meaning you negotiate. Compared to seat-priced tools like Cursor, you pay for agent usage rather than seats, which suits small teams with spiky workloads and hurts teams that run agents continuously. Because the top tier is Custom, mid-size teams should model usage before assuming it beats a flat subscription.
In short
TabTin — Open-source team harness where humans and multiple AI sub-agents share one task surface for code, docs, spreadsheets and browser research. Best for Small product teams that already run on group chat and want agents inside that workflow, Engineering teams that need auditable, rollback-able agent execution before production writes, Projects mixing coding, research, documents and spreadsheets on one shared task surface. Free to use.
What's new in TabTin
Checked 7 days agoAcross the latest 2 updates: 1 changelog entry and 1 news mention.
Code-level comparative audit of open-source team agent harnesses published
Homepage audit reviews 28 candidate GitHub open-source projects as of August 2026, narrows to 4 for deep review at fixed commits, and states TabTin is the only project to pass all checks. Self-published and not independently verified.
Competitive analysis run over 42 products, 34 scored, 3,549 reviews
A documented TabTin run scraped 42 hot AI products from watcha.cn on 2026-07-28, wrote 42 records into a shared multi-dimensional table, and produced a linked analysis report where 34 products had scores and average rating was 7.28.
Viability Score
How well maintained and how widely used is TabTin? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: September 2026
How we score →Key Features
- Main Agent decomposes a task into research, build and verification sub-agents sharing one context
- Checkpoint saved before every write, with rollback of execution state
- Human approval gate required before publish or write actions
- Repository permissions enforced per team member and connector
- 48 official application packages (Skills) selectable from the app section
- Code surface: reads repos, edits code and runs tests with visible diffs
- Document surface: drafting, rewriting and annotation with saved versions
- Spreadsheet surface: turns scattered information into filterable, countable records
- Browser surface: opens live pages and extracts results into structured data
- GitHub connector for repository reads and gated writes
- Browser connector for live web research and structured extraction
- Group chat where agents detect anomalies from messages and logs
- Automatic task creation and assignment from group-chat signals such as bug reports
- Automatic status tracking that syncs task changes back into the group chat
- Shared team resources: docs and multi-dimensional tables reuse the same research data
About TabTin
TabTin is an open-source team harness: a shared work surface where people and multiple AI agents push the same task forward together. A Main Agent splits work into research, build and verification sub-agents that inherit the same goal, team rules and delivery standards, so you stop re-explaining context every time a task changes hands. The product is downloadable (a current release is dated 9.21) and built around four concrete surfaces: a code surface that reads repos, edits code and runs tests with visible diffs; a document surface for drafting, rewriting and annotation with saved versions; a spreadsheet surface that turns scattered data into filterable, countable records; and a browser surface that opens live pages and extracts results into structured data. Shared team resources (docs and multi-dimensional tables) reuse the same research data instead of duplicating it. Collaboration runs through group chat, where agents detect anomalies from messages and logs, auto-create and assign tasks, and sync status changes back into the chat. A personal AI persona can be configured with its own rules, memory, skills and tool set; 48 official application packages (Skills) are selectable from the app section. Writes and publish actions sit behind a human approval gate with repository permissions — the part that makes agents usable in a real delivery workflow rather than a sandbox. Checkpoints are saved before changes so a run can be rolled back instead of rebuilt. TabTin targets the same ground as autonomous coding agents like Cursor or Devin but adds the team layer — checkpoints, approvals, shared resources and chat-native tracking. The primary UI and documentation are Chinese, though the site offers English, Japanese, French, German and Korean locales.
Behind the Verdict
TabTin is best understood as a team harness layered on top of agentic AI, not another chat UI. The Main Agent decomposes a task into research, build and verification sub-agents, and all three inherit the same goal, team rules and delivery standards — that context propagation is the core engineering claim, and the execution record auto-saved per stage (with a reviewable deliverable) is what makes the run auditable after the fact. Strengths worth naming: a checkpoint is saved before every write, so a bad run is rolled back rather than rebuilt; publish and write actions require explicit team confirmation, and GitHub repository write permissions are gated by a named teammate's approval; group chat closes the loop in both directions, since agents detect anomalies from messages and logs, auto-create and assign tasks, and sync status changes back to the chat. The 48 official application packages (Skills) and team connectors (GitHub, Browser) give agents something concrete to work with rather than open-ended tool use. Where it fits: small product and engineering teams that already run on group chat and mix code, documents, spreadsheets and web research in one project. Where it doesn't: solo users wanting a lightweight personal assistant, teams that need a mature third-party integration marketplace out of the box, and English-first buyers expecting full documentation parity — the scraped pages are almost entirely Chinese-language marketing copy. Access appears to be a desktop download rather than a browser product, and the developer-support page is a Prompt library for people who intend to rebuild TabTin into their own product, not a conventional API reference. Anyone who won't self-host open-source software will find the ceiling here quickly.
Researching TabTin? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas TabTin actually fits — and what changes day-one when you adopt it.
You paste a task — 'turn the hero and resource-community modules of our site into a reviewable, publishable work surface' — into the shared surface and pick PMO execution mode. The Main Agent splits it into a research sub-agent checking brand copy and mobile breakpoints across 12 product pages, a build sub-agent updating the modules with the design-system Skill, and a verification sub-agent
Outcome: You review the saved checkpoint and the auto-written execution record, approve the GitHub write against your repo permissions, and publish — with the option to roll back to checkpoint #03 if the regression fails.
Your release group chat gets a message that external imports occasionally trigger a full session refresh. The agent watching the chat spots the third identical event, writes up a reproduction path, and auto-creates task #4820 at P1 to the desktop-session owner.
Outcome: The assignee fixes it, the agent observes the fix through the release, status changes sync back into the chat, and the task closes automatically after publish.
You ask the agent to scrape 42 hot AI products from a live site into a multi-dimensional table, then draft an analysis report that cites that table. Some records lack scores, so the agent substitutes safe defaults and notes it; the report references the same 42 structured records rather than a copy.
Outcome: You get a filterable 42-record table and a versioned report in the same workspace, and you can fork the task from the 'generate report' node to extend the analysis without rerunning the scrape.
Use Cases
- Run a parallel research/build/verify agent workflow on a website redesign and review checkpoints before publish.
- Ingest a product list from a real website into a structured multi-dimensional table with scored records.
- Draft and rewrite a competitive analysis report from 42 structured records with version history.
- Let agents watch a team group chat, auto-file bug tasks from anomaly reports, and assign an owner.
- Gate GitHub repository writes behind a named teammate's approval while agents run tests.
- Convert scattered research into a filterable table that later documents cite directly.
Limitations
- The scraped evidence is almost entirely Chinese-language marketing copy; no pricing table, changelog detail or API documentation was captured, so concrete limits such as rate limits, context window size or plan gating cannot be verified from the data.
- Write and publish actions still require explicit team confirmation, and GitHub repository write permissions remain gated by human approval.
- The observable collaboration model is a main agent splitting work into research, build and verification sub-agents with checkpoints saved before every write for rollback.
as of 2026-09-21
Verification history
We have re-verified TabTin 5 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-checked, vendor evidence unchanged
- — re-checked, vendor evidence unchanged
- — re-checked, vendor evidence unchanged
- — re-checked, vendor evidence unchanged
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published TabTin tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Free
$0
Ideal for
Small teams evaluating a self-run agent harness, or teams that want agent execution records and rollback before spending anything.
What this tier adds
Starting tier: free download of the open-source TabTin harness with 48 official application packages (Skills), group chat with agents, and all four work surfaces (code, document, spreadsheet, browser) with checkpoint save and rollback before writes.
Team (credits-based)
Custom
Ideal for
Product and engineering teams running agents on real delivery work who need credit accounting, per-member allocation and a human approval gate before publish.
What this tier adds
Adds a team credit allowance tracked in organization settings, usage monitoring with remaining-credit visibility, per-member allocation and adjustment of credit usage, the human approval gate on publish/write actions, shared team resources and multi-dimensional tables, GitHub
Where the pricing makes sense
The company stage and team size where TabTin's pricing actually pencils out — and where peers do it cheaper.
TabTin prices as freemium with a free open-source download and a credits-based Team tier whose price is quoted as Custom — meaning you negotiate. Compared to seat-priced tools like Cursor, you pay for agent usage rather than seats, which suits small teams with spiky workloads and hurts teams that run agents continuously. Because the top tier is Custom, mid-size teams should model usage before assuming it beats a flat subscription.
Setup time & first value
How long it actually takes to get something useful out of TabTin — broken out by persona, not the marketing-page minute.
Engineering teams with an existing repo can expect the fastest path: download the desktop client, point a GitHub connector at the repository, set repository permissions and an approval owner, and you are running an agent task the same day. Teams adopting the shared-surface model without a connector will want a few hours to configure Skills, roles and the group chat. Teams self-hosting the source
Switching to or from TabTin
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From Cursor: keep your repo, add TabTin as the team layer so agent runs write checkpoints and route publish actions through a named approver instead of happening per-developer.
- →From Devin: move recurring research/build/verify work onto the shared task surface so sub-agents inherit team rules rather than re-deriving them per session.
- →From ad-hoc chat tools: bridge them with the group-chat workflow — agents create and assign tasks from the messages your team already posts.
- ↗To a single-player coding agent: your repos are unchanged, so you lose the checkpoint/approval trail and shared resources but keep the codebase.
- ↗To a hosted project tool: export the multi-dimensional tables and documents as the durable artifacts; agent execution records are the part that does not travel cleanly.
Integrations
Resources & Guides
Tutorials & Learning
YouTube returned 6 videos for “TabTin”, and we withheld 6: 6 could not be judged, because “TabTin” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about TabTin.
Official links
Tools that pair well with TabTin
Common stack mates teams adopt alongside TabTin, with the specific reason each pairing earns its keep.
Zhipu GLM
Zhipu GLM delivers open-source LLM models, MaaS APIs, and autonomous agents for Chinese enterprises and developers.
OpenHands
Open-source platform for autonomous cloud coding agents that fix bugs, review PRs, and automate workflows.
Gemini
Gemini is Google's multimodal AI assistant for drafting, research, and grounded answers across Gmail, Docs, and Search.
Featured Head-to-Head Comparisons
Tabtin vs Cryptohopper
These are not competing products. TabTin is an open-source team harness for software, docs, spreadsheets and browser research with a human approval gate before every write; Cryptohopper is a cloud crypto trading bot that automates orders on major exchanges and costs $24.16/mo after a 3-day trial. No budget owner shortlists both — if you need auditable agent workflows across code and documents, look at TabTin; if you need automated crypto order execution, that is Cryptohopper. Choosing between them is only a question if you happen to run both a product team and a trading account.
Tabtin vs Air Ai
These two products don't compete, and no real buyer would weigh them head-to-head. TabTin is a downloadable, freemium, open-source harness where a small product team points research/build/verification sub-agents at its own repo, docs and spreadsheets, with a checkpoint before every write and a human approval gate. Air (formerly Govini) is a contact-priced, vendor-deployed defense readiness platform whose customers are military commands and acquisition offices, judged on things like compressing Army Materiel Release from 15 months to 3 and sustaining 90% equipment readiness. Pick TabTin if you have a code/doc/research backlog and want agents inside your group chat; pick Air only if you run a defense sustainment or acquisition mission — and if you did, TabTin would never appear on your shortlist.
Tabtin vs Genspark
Pick TabTin if your problem is production-grade work with agents — code in real repos, gated writes, auditability, rollback — and you're willing to self-host an open-source harness with GitHub as your only connector. Pick Genspark if you want the opposite trade: a hosted, broad content-and-research workspace where a non-technical person spins up Super Agents, Sparkpages, slides, sheets, podcasts and video in one account, with Google Workspace, Microsoft 365, Canva and Figma wired in. TabTin gives you control and accountability; Genspark gives you breadth and zero setup. Small Chinese-language teams with engineering in the loop tilt to TabTin; marketers, researchers and no-code builders tilt hard to Genspark.
Tabtin vs Temporal Ai
These are not competing products, and a buyer should not frame a choice between them. TabTin is a shared surface where humans and AI sub-agents collaborate on code, documents, spreadsheets and browser research, with human approval gates, checkpoint rollback and GitHub connector. Temporal AI is infrastructure for durable execution: workflows and AI agents that survive crashes and retries, with native SDKs in eight languages. If you have a messy team workflow with agents in chat, TabTin fits. If you need an execution engine that keeps long-running state alive across failures, you want Temporal.
Tabtin vs Spider Cloud
These are not competitors and shouldn't be shortlisted against each other. Pick TabTin if you're a small product team that wants humans and multiple AI sub-agents pushing one auditable task — code, docs, sheets, browser research — through a self-hosted surface with approval gates and rollback. Pick Spider Cloud if your actual problem is that the data you need lives on websites with no API, and you want one key that returns rendered markdown, full-site crawls, or search results into your own agents and RAG pipelines. If you're a team of two, running both is plausible but they solve different layers of the stack.
Tabtin vs Cognition Ai
These are not substitutes. Devin (Cognition AI) is a managed enterprise engineer: you point it at a large production repo and it plans, codes, tests, opens PRs, auto-triages bugs, and clears vulnerability backlogs, backed by FedRAMP High In-Process and a $10M productivity guarantee. TabTin is an open-source harness you download and self-host so a small team's people and several sub-agents share one task surface across code, docs, spreadsheets, and browser research, with a checkpoint-rollback and human approval gate on every write. Pick Devin if you have review capacity, a big codebase, and a compliance story; pick TabTin if your problem is broader than code, you want no vendor contract, and you're willing to run the software yourself.
Alternatives to TabTin
View all
Frequently Asked Questions
Categories
Best-of guides
Used TabTin? Help shape our editorial sentiment research.