01 / MEMORY
Memory that stays put.
It remembers the project across threads and sessions — and only inside the room it learned it. Tag it elsewhere and it cannot read what it was never given.
Open source Claude Tag alternative in Feishu.
Support all agents, all models.
Open sourced
Free · no card
Trusted by users from
OpenAI
GoogleCoinbaseByteDanceGUAZI.COM
GoogleCoinbaseByteDanceGUAZI.COM
@OpenTag it, and it’s handled.
Drop a to-do into any thread and get the finished work back in the same chat.
Product Eng
Priya 4:09 PM
@OpenTag pull what merged since 4.1 and draft the 4.2 release notes.
OpenTag Bot 4:12 PM
Diffed
4 features, 9 fixes, 2 breaking:
•
• Config schema bumped to v2
• Webhook retries now exponential
• Rate-limit headers renamed
Two need a migration note. Drafted the whole thing in your house format — want me to post it?
v4.1.0…v4.2.0 — 63 commits.4 features, 9 fixes, 2 breaking:
•
--legacy-auth removed• Config schema bumped to v2
• Webhook retries now exponential
• Rate-limit headers renamed
Two need a migration note. Drafted the whole thing in your house format — want me to post it?
🚀 4🙏 2
Customer Support
Wang Lei 4:09 PM
@OpenTag group this week’s complaints by theme, flag anything raised 3+ times.
OpenTag Bot 4:12 PM
Read 214 tickets, Aug 4–10.
11 themes. Three cleared the threshold:
• Login loop on SSO — 19
• Export times out >10k rows — 12
• Billing page 404 after refund — 7
SSO is up 4× on last week. Exported to Base, owners @-ed.
11 themes. Three cleared the threshold:
• Login loop on SSO — 19
• Export times out >10k rows — 12
• Billing page 404 after refund — 7
SSO is up 4× on last week. Exported to Base, owners @-ed.
👍 6🎉 3
Marketing
Alina 4:10 PM
@OpenTag the site still says “beta” everywhere and launch is Monday — can you sweep it?
OpenTag Bot 4:12 PM
Swept 6 pages.
• 14 “beta” strings replaced
• Pricing date set to Aug 17
• Two stale screenshots flagged
• OG image regenerated
The pricing page also still links to the old changelog. PR is up with before/after shots.
• 14 “beta” strings replaced
• Pricing date set to Aug 17
• Two stale screenshots flagged
• OG image regenerated
The pricing page also still links to the old changelog. PR is up with before/after shots.
👀 3🔥 2
Data
Zhao Min 4:10 PM
@OpenTag pull last week’s D1 and D7 retention, split by channel.
OpenTag Bot 4:12 PM
Scanned 2.1M events, Aug 4–10.
D1 41.2% ▲ 3.1
D7 18.7% ▼ 0.8
D14 12.4% ▲ 1.2
D30 9.1% ▲ 0.4
Split by 5 channels. Paid social is the weak one — D7 there is 11.2%, well under the blended number.
Charts posted to Files.
D1 41.2% ▲ 3.1
D7 18.7% ▼ 0.8
D14 12.4% ▲ 1.2
D30 9.1% ▲ 0.4
Split by 5 channels. Paid social is the weak one — D7 there is 11.2%, well under the blended number.
Charts posted to Files.
📊 4🌈 2
Ops & Infra
Sam 4:02 PM
@OpenTag the staging deploy failed twice this morning — what happened?
OpenTag Bot 4:12 PM
OOM kill in the build step,
Both runs died at the same step:
• 09:14 — build, 4m 02s
• 10:31 — build, 3m 58s
Tuesday’s config change capped the Node heap at 2 GB and the bundle crossed it.
PR up to restore 4 GB.
exit 137.Both runs died at the same step:
• 09:14 — build, 4m 02s
• 10:31 — build, 3m 58s
Tuesday’s config change capped the Node heap at 2 GB and the bundle crossed it.
PR up to restore 4 GB.
✅ 3🙌 1
Bring your own everything
You already pay for this. Pay for it once.
OpenTag never touches your tokens. It drives the agent already on your machine.
On subscriptions: consumer plans from Anthropic and OpenAI are written for their own apps. Third-party use is governed by each provider's terms. API keys are the fully supported path.
What you get
A teammate you can read the source of.
02 / PERMISSIONS
Permissions, per channel.
Tools are scoped to the channel, not shared globally. Engineering reaches GitHub; legal only sees contracts. In a DM it falls back to your own account and connectors.
03 / INTEGRATIONS
Connect your stack.
GitHub, GitLab, Notion, Google Calendar and more — your agent reaches the tools your team already uses.
04 / SKILLS
Skills in plain files.
A skill is a markdown file. Write it, commit it, review it in a PR.
Social proof
What builders say.
It finally feels like a build tool that works the way I want.
Running six agents in parallel and reviewing every diff in one place. This is the workflow I wanted.
As a designer, the UI is the first thing I judge. This one actually holds up.
It's like adding five engineers to a two-person startup, overnight.
Agents triage my repo's issues and open PRs while I sleep. For a solo maintainer, that's huge.
Beautiful UI, and it handles all the git worktree stuff for you.
Compare
The same job, without the middleman.
| OpenTag | Claude Tag | Ogment | Adapt | Viktor | |
|---|---|---|---|---|---|
| Open source | Apache 2.0 | — | — | ||
| Runtime support | Claude Code, Codex, Cursor, Kimi, opencode | Anthropic's own | — | — | Its own agent |
| Bring your own plan | Any API key, Claude & Codex plans | DMs only | — | — | — |
| Run on your own machine | — | ||||
| IM integration | Feishu / Lark · Slack soon | Slack | Slack | Slack | Slack, Teams |
| Option to self-host | Docker | — | |||
| Pricing model | Free · you pay your provider | Usage balance | Credits | Usage credits | Workspace credits |
Checked against each vendor’s own docs and pricing pages, August 2026. — means the vendor doesn’t state it publicly, not that the answer is no. Something wrong? Open an issue and we’ll correct it.
Just tag it.
The future of work is humans and agents building side by side.
Questions
Can I really use my Claude or ChatGPT subscription?
OpenTag supports signing in with a Claude or Codex plan as well as with an API key. Worth knowing: consumer subscriptions are sold for use in the provider's own apps, and whether third-party access is permitted is governed by each provider's terms, which change. API keys are the unambiguous path and what we recommend for a shared workspace. Read your plan's terms and decide for your own team.
How is this different from Claude Tag?
Claude Tag is Anthropic's own Slack integration: polished, hosted, Anthropic models only, priced per seat. OpenTag is the open equivalent you run yourself, in Feishu, on whichever agent and model you already pay for. If you are happy on Anthropic models in Slack and want it managed for you, Claude Tag is a good product. Choose OpenTag when you want Feishu, the source, or your own plan.
What does it cost?
The software is free and Apache 2.0 licensed. You pay your model provider directly at their rates, plus whatever your own hosting costs. No credit system and no markup on inference.
Where does our data go?
To your infrastructure and to whichever model provider you configure. Messages, memory, and logs sit in a database you run. There is no OpenTag-operated server in the path, which is the practical difference from every hosted option on this page.
Do I need to be technical to set it up?
Whoever sets it up needs to create a Feishu bot and run the agent on a machine they control. After that, using it is just tagging a bot in a thread, so the rest of the team needs no setup at all.
Does it work in Lark as well as Feishu?
Yes. They are the same platform under two names, so one bot configuration covers both. Slack is not supported today.
Can we contribute?
Please do. Issues, connectors, and skills are all welcome — start at the GitHub repo.