2026-06-18

Alibaba Supplier Outreach with OpenClaw (2026)

Alibaba supplier outreach with OpenClaw: find suppliers via LaunchFast, send outreach, check replies, and negotiate in one tracked workflow.

2026-06-17

Alibaba Supplier Outreach with OpenClaw (2026)

A practical tutorial for the @blockchainhb/alibaba-supplier-outreach skill on ClawHub: find Alibaba suppliers through LaunchFast, send personalized outreach, track replies, and run negotiation rounds — all from a single OpenClaw agent that lives in your Chrome session.

Sourcing a product on Alibaba in 2026 still takes most Amazon FBA sellers the same three days it took in 2019. You search for a product, copy ten supplier names into a spreadsheet, draft the same "Hi, I am an Amazon FBA seller" message ten times with the company name swapped, paste each one into Alibaba's contact form, and then check the Message Center every morning for a week to see who replied. Multiply that across three product lines and you have lost a quarter to logistics that should be a chat message.

The alibaba-supplier-outreach skill on ClawHub turns the whole loop into one tracked workflow. The skill handles supplier discovery through LaunchFast, sends messages through Chrome automation so it uses your real logged-in Alibaba account, checks the Message Center for replies, and drafts stage-aware negotiation replies — all from a single OpenClaw agent that writes its conversation log to ~/.claude/supplier-conversations/ as it goes. 1,400 installs on ClawHub, v1.0.0, by @blockchainhb. This article walks through a real sourcing run from "find suppliers for silicone spatula" to "I have three replies and a counter offer ready," including the LaunchFast call, the Chrome automation steps, the message template, and the four negotiation stages.

What the alibaba-supplier-outreach skill does

The skill is a workflow glue layer between three tools you already use: LaunchFast (supplier discovery), Alibaba.com (the supplier directory and inbox), and Claude in Chrome (the browser automation that sends and reads messages for you). It does not scrape Alibaba directly — it uses your real logged-in session, which means it inherits Alibaba's rate limits and your account reputation.

It ships with three modes that match the three jobs an FBA seller actually has:

Mode What it does When you run it
OUTREACH Find suppliers via LaunchFast, draft personalized messages, send via Chrome Once per new product line
CHECK REPLIES Open the Alibaba Message Center, extract replies, summarize each Daily until 80% of suppliers have replied
NEGOTIATE Read the thread, determine the negotiation stage, draft a stage-aware reply Once per supplier per round

A quick note on the safety story before you install it: the OpenClaw security audit on ClawHub reports this skill as "Suspicious" — the same status it gets for any skill that automates Chrome and writes to your local filesystem. That is not a malware flag — it is a "use with extra caution, read the skill source before installing" flag. The skill source is open on ClawHub and the SKILL.md is downloadable as markdown. Read it once before you install, the same way you would read a package.json before npm install. The pattern it follows is the standard one for OpenClaw browser skills: it uses mcp__claude-in-chrome__* tools (which only run in Chrome tabs you already have open) and writes its logs to a path you can inspect.

Why Alibaba outreach is broken without automation

The cost of manual outreach is not the typing. It is the state.

A real sourcing run for a new product line has three concurrent threads: who you contacted, what you said, what they replied with. Without a tracker, every reply becomes a memory test: did I already reply to this supplier, did I commit to that price, what did they say about the MOQ? Multiply that across 10 suppliers and 4 negotiation rounds and the whole process breaks down around day three.

The pattern the skill implements is the standard agent workflow for stateful business processes: a research phase, an outreach phase with user approval gates, a polling phase for replies, and a negotiation phase that reads prior context before drafting the next message. The agent does the typing; the human approves every send and makes every pricing decision. That approval gate is the difference between an automation tool and a spam tool.

The other thing that breaks without automation is the signal-to-noise ratio. Alibaba's default search returns every supplier that paid for placement, not every supplier you should contact. Gold Supplier + Trade Assurance filters cut the list by ~80%. LaunchFast scores the rest. Without those filters you spend your one intro message on a supplier who will not reply because their MOQ is 10,000 units and your target is 500.

Prerequisites

You need four things before the first run.

  1. OpenClaw with the agent runtime running. If you are on GolemWorkers this is already done — the agent is hosted and your skill registry is global.
  2. Chrome with the Claude in Chrome extension installed and connected. The skill uses mcp__claude-in-chrome__* tools. Without the extension, every Chrome call returns "tool not available" and the workflow stops at step 4.
  3. An Alibaba.com account, logged in. Open https://www.alibaba.com in Chrome and complete the login before the agent session. The skill does not handle Alibaba auth — it uses your session cookies.
  4. A LaunchFast account with supplier research access. The skill calls mcp__launchfast__supplier_research to find suppliers. LaunchFast is paid, billed per query, and has a free tier that covers 5-10 products/month for testing.

You do not need an Alibaba Gold Supplier or Trade Assurance badge yourself — those are supplier-side credentials. You do not need an API key for the skill itself — the install command pulls the latest from ClawHub.

# In the worker's env (or terminal if you run OpenClaw locally)
          openclaw skills install alibaba-supplier-outreach

          # Restart the agent so the new skill shows up in the tool list
          openclaw restart

After the restart, the skill shows up as alibaba-supplier-outreach with the three modes above as separate intents. Test that the launchfast tool is wired up with a one-liner:

openclaw tools list | grep launchfast
          # Expected output: mcp__launchfast__supplier_research (registered)

If launchfast is missing, install the MCP server separately:

openclaw mcp install launchfast
          export LAUNCHFAST_API_KEY=*** ~/.secrets/launchfast-token)

Step 1. Run the LaunchFast supplier research

Start the agent and tell it what you are sourcing. The skill's first job is to ask you the four parameters it cannot guess — product keyword, target price, target quantity, your name and experience — then call LaunchFast with filters that cut the noise.

You: find suppliers for silicone spatula, target $1.50 landed, 500 units,
          I am Max, 2 years on Amazon
          Agent: Got it. Running LaunchFast research now with Gold Supplier + Trade
          Assurance filters, max 10 results.

Behind the scenes the skill calls:

mcp__launchfast__supplier_research \
          keyword="silicone spatula" \
          goldSupplierOnly=true \
          tradeAssuranceOnly=true \
          maxResults=10

LaunchFast returns a scored list. The skill presents it as a table you can actually act on:

# Supplier Score Price range MOQ Years Trust
1 Sheng Jie (Dongguan) Silicone 76 $1.15–$1.25 100 15 Gold, TA, Assessed
2 Ningbo Kebon Houseware 71 $1.20–$1.35 200 9 Gold, TA
3 Dongguan Wellbest Silicone 68 $1.05–$1.20 500 12 Gold
4 Yiwu Yuxi Import & Export 62 $0.95–$1.15 1000 6 TA only
5 Shenzhen Sunrising Houseware 58 $1.30–$1.45 300 8 Gold

Two things to look at before you pick: score (composite of trust badges, years, response rate) and MOQ vs your target quantity. Supplier 4 has the lowest price but a 1,000-unit MOQ against your 500-unit target — that is a structural mismatch, not a price problem. Cross them off and pick from the top three.

Step 2. Pick suppliers and approve the outreach style

Tell the agent which suppliers to contact (by number) and which message style you want:

You: contact 1, 2, 3. Style A — auto-generate optimized quote request.
          Agent: Drafting three personalized messages using each supplier's name,
          years in business, and your target numbers. One moment.

The skill's Style A is a six-part psychological outreach formula. It works because it does the opposite of what every spam outreach message does:

  1. Name the supplier specifically — company name, years in business, the verification badge that made you pick them. Signals "I researched you" not "I scraped 200 names."
  2. State buyer credibility — Amazon FBA seller, years on the platform, scaling this product line.
  3. Anchor with real numbers — target quantity + target price. Never "how much for samples" (they will quote retail).
  4. Soft urgency — "evaluating 2-3 suppliers this week." No fake deadlines.
  5. Three specific questions — price at your quantity, lead time, private label capability. Three is the right number: one is too few to look serious, five feels like a form.
  6. Warm close — invite a call if easier. Many senior Alibaba sellers prefer voice.

Here is what the agent drafts for supplier 1:

Hi Sheng Jie Team,

          I came across your company while sourcing silicone spatula suppliers —
          15 years of experience and your Assessed Supplier verification really
          stood out.

          I'm an Amazon FBA seller scaling my silicone kitchen line (2 years on
          Amazon) and looking to establish a reliable long-term manufacturing
          partner.

          I'm evaluating 2-3 suppliers this week and would love some details:

          1. Best price for silicone spatula at 500 units? (targeting ~$1.50/unit
          landed)
          2. Standard production lead time for that quantity?
          3. Can you do custom private label packaging (logo + custom colors)?

          Ready to place a trial order within 2-3 weeks if we're aligned.
          Happy to jump on a call if that's easier.

          Best,
          Max

Read the message before approving. The skill always shows you the draft and waits for a yes/no. Approve all three (or edit one — the skill takes edits).

Step 3. Send through your logged-in Chrome session

Once you approve, the skill opens your real Alibaba session in Chrome and sends. This is the part that requires the Claude in Chrome extension — the skill cannot bypass Alibaba's contact form because Alibaba blocks bots that do not use a real browser fingerprint.

The skill navigates to each supplier's contact form using the URL pattern:

https://www.alibaba.com/trade/search?tab=supplier&SearchText=<encoded-company-name>

Where the company name is URL-encoded: Sheng Jie (Dongguan) Silicone Rubber becomes Sheng+Jie+Dongguan+Silicone+Rubber.

It then walks through the contact form: click the "Contact Supplier" button, wait for the message page to load, find the message textarea by its visible label, type your approved message, click "Send Inquiry Now," and wait for Alibaba's success page (feedbackInquirySucess.htm is the URL that confirms the inquiry landed).

Important rules the skill enforces:

  • Screenshot before and after every form interaction. Alibaba's contact page shifts its layout every few months. The skill takes a screenshot before clicking anything so it can verify the button is still where it expects.
  • Verify the "To:" field shows the right supplier. Alibaba sometimes shows a related supplier instead of the one you searched for. If the recipient is wrong, stop and do not send.
  • One supplier per Chrome tab. The skill does not switch tabs mid-message; if you contact 3 suppliers, it uses 3 tabs.

Three inquiries take about 90 seconds end-to-end because the Chrome extension handles the typing and clicking faster than a human could.

Step 4. Save the conversation logs locally

The moment a message sends successfully, the skill writes the conversation to disk. The structure lives under ~/.claude/supplier-conversations/ and is what makes the rest of the workflow possible — without this, the agent has no memory of who it contacted.

~/.claude/supplier-conversations/
          ├── index.md                                    # master list of all suppliers
          └── sheng-jie-dongguan-silicone-rubber/
          └── conversation.md                         # full thread log for this supplier

The index.md is a one-row-per-supplier summary you can read in 10 seconds to see the state of every active sourcing run:

# Supplier Negotiations

          | Supplier | Product | Stage | Their Price | Target | Last Contact |
          |----------|---------|-------|-------------|--------|--------------|
          | Sheng Jie | silicone spatula | outreach_sent | — | $1.50 | 2026-06-17 |
          | Ningbo Kebon | silicone spatula | outreach_sent | — | $1.50 | 2026-06-17 |
          | Dongguan Wellbest | silicone spatula | outreach_sent | — | $1.50 | 2026-06-17 |

The per-supplier conversation.md records every send and reply with timestamps. The skill updates this file immediately after sending — never batched — because the cost of losing state on a live negotiation is much higher than the cost of an extra write call.

Step 5. Check replies (MODE 2)

Most Alibaba suppliers reply within 24-72 hours. The skill's second mode exists to read those replies without making you log into Alibaba five times a day.

You: check replies
          Agent: Opening Message Center. Reading conversation list. Three unread
          conversations found.

The agent navigates to https://message.alibaba.com/message/messenger.htm#/, reads the left panel for unread indicators, opens each unread conversation in turn, extracts the reply text, your target price from the conversation log, and presents a structured summary:

## Reply from Sheng Jie (Dongguan) Silicone
          Received: 2026-06-18 14:32

          Their message:
          > "Dear Max, thank you for your inquiry. For 500 units silicone spatula
          >  with custom logo, our best price is $1.35/unit FOB Shenzhen. Lead time
          >  25 days after deposit. We can do custom packaging. Free samples available,
          >  you pay shipping. Please let us know how to proceed."

          Key data:
          - Their price: $1.35      |  Your target: $1.50     |  Gap: 11% above target
          - Lead time: 25 days
          - MOQ: 500 units (matches your target)

          Negotiation stage: initial_reply
          Suggested next step: counter at $1.25 (below their quote, above your
          target to leave room) and ask about sample cost + shipping estimate.
          Want me to draft and send a reply? (yes / show me options / no)

That summary is the agent's job. The decision to send the counter is yours.

Step 6. Negotiate with stage-aware replies (MODE 3)

When you say yes, the skill switches to MODE 3 — NEGOTIATE. It re-opens the conversation in Chrome, reads the full thread, loads conversation.md to determine which stage you are in, and drafts a reply tailored to that stage.

Stage Signal in the thread Reply strategy
1 — First reply They responded to your initial inquiry Thank, counter price, ask about samples
2 — Counter exchange They gave a price, you need to push Propose compromise, offer value levers (faster payment, larger order)
3 — Closing Price agreed or close Confirm terms, request proforma invoice, ask about sample timing
4 — Ongoing Established relationship Direct, brief, reference previous order

For Sheng Jie, the agent is in stage 1 and drafts:

Thank you so much for the quick reply — really appreciate the detail.

          $1.35 is a bit above what I had budgeted for the trial. Could you do
          $1.25/unit at 500 units if I commit to a 30% deposit within 48 hours
          of agreeing on terms? Lead time of 25 days works perfectly on my side.

          For samples — happy to pay shipping. Could you confirm the sample cost
          (just the spatula, no custom logo yet) and the courier options to a US
          address?

          If we can align on $1.25, I'm ready to move to PI this week and place
          the trial order in the next 2-3 weeks.

          Best,
          Max

The skill shows you the draft and waits for approval. If you approve, it pastes the message into the open conversation in the Message Center, clicks Send, and updates conversation.md with a new entry:

### 2026-06-18 16:05 SENT — Counter Offer
          [message text above]

          - Their last offer: $1.35
          - Our counter: $1.25 (8% below their offer, 17% below target)
          - Levers offered: 30% deposit within 48 hours
          - Sample ask: yes

You repeat this loop — counter, reply, counter — until the price is in your zone or you decide to walk. Most sourcing runs take 3-5 rounds over 7-10 days.

End-to-end: source a silicone spatula in one afternoon

Here is the full timeline a real sourcing run takes with the skill installed:

Time What you do What the agent does
0:00 Type "find suppliers for silicone spatula, target $1.50, 500 units" Calls LaunchFast, returns top 5 table
0:05 Pick 1, 2, 3 and approve Style A Drafts three personalized messages
0:10 Approve all three messages Sends through Chrome, saves 3 conversation logs
Day 1–3 (nothing) —
Day 3 morning Type "check replies" Opens Message Center, returns 2/3 replies with structured summary
Day 3 afternoon Approve counter for Sheng Jie, decline to reply to Wellbest Sends counter for Sheng Jie, marks Wellbest as dead in index.md
Day 4–7 (nothing) —
Day 7 Type "check replies" Returns Sheng Jie's response at $1.30
Day 7 Approve "accept $1.30, request PI + samples" Sends closing message, updates stage to closing
Day 10 Receive sample + PI from Sheng Jie —
Day 10 Type "place trial order" Marks order placed in conversation.md

Total human time: ~40 minutes over 10 days. The same run manually takes most FBA sellers 6-10 hours of clicking and copy-pasting.

Common pitfalls

The skill enforces good defaults, but there are four traps that bite every new user.

  • Logging out of Alibaba between sessions. The skill uses your Chrome session cookies. If you close Chrome or Alibaba expires your session, the next "Send Inquiry" silently fails because the contact form redirects to login. Always verify you are still logged in before starting a session.
  • Picking suppliers with MOQ above your target quantity. Supplier 4 in the example table has the lowest unit price but a 1,000-unit MOQ. A low price with an unreachable MOQ is not a deal — it is a teaser. Filter by MOQ ≤ your target before reading the price column.
  • Letting the agent auto-send without showing you the draft. The skill defaults to "show draft, wait for approval" — never change this to "auto-send." A typo in the target price or a misread of the supplier's name becomes a permanent part of the negotiation thread and is hard to walk back without losing face.
  • Forgetting to update index.md when a supplier goes cold. If a supplier has not replied in 7 days, mark them as dead in the index so the agent stops suggesting follow-ups. The skill does this for replies but not for silence — that part is on you.

What to automate next

Once supplier outreach is running, three related workflows fit the same pattern:

  • Sample tracking. A second skill can read conversation.md files, extract when each supplier promised to ship samples, and send you a Telegram ping if the sample window expires without a tracking number.
  • PI → invoice reconciliation. When the proforma invoice arrives, a parser can extract line items, cross-check them against your negotiated terms, and flag any drift (unit price changed, MOQ changed, lead time changed) before you wire the deposit.
  • Quality control photos. Many suppliers send QC photos before shipment. A skill can watch your Gmail for attachments matching the supplier's name + "QC" + image MIME type, drop them into a qc-photos/<supplier-slug>/ folder, and notify you in Telegram.

All three compose on top of the same ~/.claude/supplier-conversations/ memory the outreach skill writes, so the state stays consistent across skills.

FAQ

What is alibaba supplier outreach automation?

Alibaba supplier outreach automation is the use of an AI agent to find suppliers, draft and send the initial inquiry, check the Message Center for replies, and run negotiation rounds without a human doing each step manually. The @blockchainhb/alibaba-supplier-outreach skill on ClawHub is one such automation — it combines LaunchFast for supplier discovery with Claude in Chrome for browser automation and writes its state to ~/.claude/supplier-conversations/.

How do I find Alibaba suppliers for a product?

The fastest way in 2026 is LaunchFast filtered to Gold Supplier + Trade Assurance, which cuts the default Alibaba search results by roughly 80% and gives you a scored shortlist. The outreach skill wraps this in a one-line chat command: "find suppliers for [product]." Without LaunchFast, you can use Alibaba's native filters but you will spend more time weeding out suppliers with unrealistic MOQs or no trade assurance.

How long does Alibaba outreach take with automation?

A typical first outreach round (find 5-10 suppliers, send 3-5 personalized inquiries) takes 10-15 minutes of human time once you have Alibaba logged in and Chrome with the Claude extension ready. Without automation, the same round takes 60-90 minutes of clicking and copy-pasting. Negotiation rounds (check replies, draft counter, send reply) take 2-3 minutes each, mostly waiting for you to approve the draft.

Is it safe to automate Alibaba messages?

The skill uses your real Chrome session via the official Claude in Chrome extension — it does not bypass Alibaba's contact form. That means you stay within Alibaba's terms of service, you keep your account reputation, and Alibaba's anti-bot signals do not flag you. The skill's OpenClaw security audit reports "Suspicious" because it automates Chrome and writes to local files; that is a "review the source before installing" flag, not a malware flag. Read the SKILL.md before installing, the same way you would read a package.json before npm install.

Can I customize the outreach message template?

Yes. The skill ships with a Style A (auto-generated, formula-driven) and a Style B (you write your own). To change the formula or add a third style, edit the skill source at ~/.openclaw/skills/alibaba-supplier-outreach/SKILL.md — the message template is in the section titled "Psychological Outreach Formula." Changes take effect on the next agent restart.

What does Alibaba supplier outreach cost?

The skill itself is free on ClawHub (open source, MIT-0 license). LaunchFast has a free tier of 5-10 supplier research queries per month, paid plans start around $49/month for higher volume. Claude in Chrome is included with a Claude subscription. The actual Alibaba messages are free (your account, your messages). Realistic cost for a sourcing run that contacts 5 suppliers and runs 3 negotiation rounds: $0 in tool costs if you stay under the LaunchFast free tier, $5-15 if you do not.

What if a supplier asks for a phone call?

The skill drafts the warm close ("Happy to jump on a call if that's easier") but does not place calls itself. If a supplier replies asking for a call, the agent logs it in conversation.md with stage call_requested and surfaces it in your daily summary. You schedule and take the call yourself; after the call, dictate the outcome to the agent ("Sheng Jie agreed to $1.27 if I commit to 1,000 units across two orders") and the skill updates the stage and the next reply draft accordingly.

Related articles