Clash Notion, Figma & Miro Workflow Setup Guide 2026
Why Notion, Figma, and Miro workflows need disciplined Clash split routing
Notion, Figma, and Miro are often treated as ordinary browser tabs, but a modern collaboration session is more complicated than loading one webpage. A Notion workspace may combine authentication, editor requests, websocket connections, integration calls, image previews, and large attachments. Figma separates interactive design traffic from fonts, thumbnails, plugin resources, and heavy file assets. Miro adds real-time board synchronization, cursor updates, embedded media, templates, and export downloads. When all of these connections inherit one accidental route, the visible symptom may be “the app is slow,” even though only one traffic family is actually failing.
This is where a Clash productivity workflow becomes useful. Instead of sending the entire computer through a global proxy, you can create intentional policy groups for collaboration services while allowing local websites, banking portals, company intranet pages, and regional services to use DIRECT. The result is not simply a faster connection. It is a setup that is easier to inspect, reproduce, and repair when a login page works but a board refuses to synchronize or a design file loads without its fonts.
The objective of this guide is a practical Clash Notion Figma Miro setup for 2026 using Clash Verge Rev, Mihomo, or another client that supports standard Clash configuration. The exact menu labels differ between desktop and mobile clients, but the configuration principles remain the same: identify traffic, assign meaningful groups, place specific rules before broad rules, keep DNS behavior consistent, and validate each application separately.
Map the traffic before writing Clash rules
A reliable policy starts with observation rather than guesses. Open Notion, Figma, and Miro one at a time, perform a normal task, and watch the connection panel in your Clash client. For Notion, open a database, edit several blocks, upload a small attachment, and reload the workspace after a short sleep period. For Figma, open a large team file, zoom across several frames, enable or inspect a font, and export a small image. For Miro, load a board with images or documents, move objects, invite a test collaborator if appropriate, and observe whether board updates continue after the tab has been idle.
Record the hostnames, the selected policy, the connection result, and the approximate time. This small log is more valuable than a collection of forum snippets because it describes your workspace, your DNS environment, and your network. A hostname that appears during login may not appear during a board edit, while a signed storage hostname may exist for only a few minutes. If you write rules only for the obvious homepage, the application can still fail during the most important operation.
Notion traffic buckets
Notion commonly involves the main workspace domain, authentication endpoints, API requests, websocket-style synchronization, and attachment delivery. Domains such as notion.so, www.notion.so, api.notion.com, notion.site, and notionusercontent.com may appear in different roles, but you should verify them against your own logs. Attachments can also be served through signed object-storage URLs, meaning the final hostname may belong to an AWS S3 or CDN domain rather than a Notion-owned domain.
That distinction matters. The editor may feel responsive when small JSON requests use a stable route, while a PDF or image download stalls on a congested exit. Conversely, routing every storage domain through a broad proxy group may send unrelated cloud traffic through the same policy. Start with the smallest useful domain set, then expand it only when an observed request clearly belongs to the Notion workflow.
Figma traffic buckets
Figma has at least two practical traffic shapes. The first is interactive application traffic: authentication, file metadata, comments, collaboration events, and canvas state. The second is asset traffic: thumbnails, fonts, plugin packages, image resources, and large file data. A rule for only figma.com may cover the application shell but leave an asset CDN on a different route. That can produce a blank canvas, missing fonts, or a file that appears to load forever.
Do not assume every Figma-related hostname should use the same group. Interactive traffic benefits from a low-latency, stable selector, while large assets benefit from a route with consistent throughput. In many environments, using one well-tested group for both is simpler and more stable than aggressively splitting them. Split only after logs show a real difference between control traffic and asset delivery.
Miro traffic buckets
Miro boards depend heavily on real-time synchronization. The board shell, account services, websocket connections, media previews, templates, and export operations may not share one hostname. If a board opens but object movements appear delayed, inspect the long-lived connections rather than changing the browser first. If editing works but an export fails, investigate the download host and its policy separately.
For all three services, remember that a browser can reuse existing connections and cached resources. Test in a clean tab, after signing out and in when safe, or after clearing only the relevant site data. Otherwise, you may believe a rule works simply because the browser is displaying an old cached shell.
Build policy groups that match daily work
A useful Clash configuration separates selection from classification. The proxy-groups section decides which exit or fallback behavior is available, while the rules section decides which traffic uses that group. Naming groups by purpose makes later troubleshooting much easier than names such as “Node 1” or “Fast VPN.” A practical desktop arrangement might include a general proxy selector, a collaboration selector, a fallback group, and a direct policy.
proxy-groups:
- name: COLLABORATION
type: select
proxies:
- AUTO-STABLE
- MANUAL-PREFERRED
- DIRECT
- name: AUTO-STABLE
type: url-test
url: https://www.gstatic.com/generate_204
interval: 300
tolerance: 80
proxies:
- relay-a
- relay-b
- relay-c
- name: LOCAL
type: select
proxies:
- DIRECT
- COLLABORATION
The names above are examples, not universal requirements. Replace the proxy names with entries that actually exist in your profile. A url-test result measures reachability to its test URL, not the exact quality of a Notion websocket or a Figma file download. Use it as a baseline, then verify real applications. A manually selected group is often better for an important presentation or design review because it avoids an automatic switch in the middle of a live collaboration session.
Keep a direct option available when your network policy and local requirements allow it. Some organizations require all work traffic to follow a managed gateway, and some services may block unexpected exit changes. Clash should be configured in accordance with local law, workplace policy, and the terms that apply to your accounts. The purpose of split routing is predictable traffic management, not hiding activity from an authorized administrator.
Write specific rules before broad rules
Clash evaluates rules from top to bottom. A specific domain rule must appear before a broad geographic rule, a catch-all proxy rule, or a final fallback. If DOMAIN-SUFFIX,notion.so,COLLABORATION appears after a broad GEOIP,CN,DIRECT line, the request may be classified before it ever reaches the intended Notion policy. Rule order is therefore part of the design, not a cosmetic detail.
rules:
- DOMAIN-SUFFIX,notion.so,COLLABORATION
- DOMAIN-SUFFIX,notion.site,COLLABORATION
- DOMAIN-SUFFIX,notionusercontent.com,COLLABORATION
- DOMAIN-SUFFIX,figma.com,COLLABORATION
- DOMAIN-SUFFIX,figmausercontent.com,COLLABORATION
- DOMAIN-SUFFIX,miro.com,COLLABORATION
- DOMAIN-SUFFIX,mirostatic.com,COLLABORATION
- GEOSITE,private,DIRECT
- GEOIP,private,DIRECT,no-resolve
- MATCH,PROXY
Use DOMAIN-SUFFIX when the service legitimately uses subdomains, and use DOMAIN when you want to target one exact hostname. Avoid adding broad suffixes such as amazonaws.com simply because an attachment happened to use S3. That could capture unrelated developer tools, software updates, or company services. When an asset is delivered through a signed CDN hostname, add it only after confirming its role and understanding the possible scope.
Separate app policies only when evidence supports them
Many users begin with three groups: NOTION, FIGMA, and MIRO. This is easy to understand and gives each service an independent switch. It also creates more maintenance. If all three services behave well on the same route, a single COLLABORATION group reduces accidental differences. If Figma assets require a high-throughput route while Miro synchronization needs low latency, create a separate DESIGN-ASSETS group only after testing confirms the benefit.
Use rule providers when your profile already supports them and when the provider is maintained. A remote rule set can reduce local editing, but it introduces another dependency and may change classification without notice. For a small personal workflow, explicit local rules are easier to audit. For a team-managed configuration, a reviewed rule provider can make updates more consistent, provided its source and update behavior are trusted.
Align DNS and TUN settings with the routing model
Rules cannot help if DNS resolution sends you to an unexpected address or if the application bypasses the Clash capture path. DNS, system proxy mode, and TUN mode should be considered together. In a browser-only setup, system proxy mode may be enough for applications that respect HTTP and SOCKS proxy settings. Notion, Figma, Miro, desktop helpers, and background processes may not all obey those settings, so TUN mode is often more consistent when your client and operating system support it.
Enable TUN only after confirming that the required driver or system permission is available. On Windows, a Mihomo-based client may need a Wintun component and administrator approval. On macOS, the client may request a network extension permission. On Linux, the interface and routing permissions vary by distribution. Android uses its VPN service rather than a desktop TUN workflow, while iOS support depends on the specific client. If TUN starts but local devices, printers, or corporate resources stop working, review bypass and private-network rules instead of assuming the proxy nodes are defective.
For split DNS, choose a design that keeps policy decisions and resolution behavior compatible. A fake-IP mode can make domain-based routing predictable for applications that tolerate synthetic addresses, but some software, local discovery tools, and security products expect real answers. If a collaboration application displays a login loop, certificate warning, or failed websocket after enabling fake-IP, test the relevant domain under redirection-host or another supported mode. Do not change every DNS option simultaneously; record one working baseline first.
A basic Mihomo DNS section might resemble the following, but the correct nameserver addresses depend on your network and provider:
dns:
enable: true
enhanced-mode: fake-ip
fake-ip-range: 198.18.0.1/16
nameserver:
- https://1.1.1.1/dns-query
- https://8.8.8.8/dns-query
fallback:
- https://dns.quad9.net/dns-query
fallback-filter:
geoip: true
This example is not a promise that public resolvers are suitable for every workplace. Use approved resolvers where required, and consider whether encrypted DNS endpoints themselves need a defined proxy policy. If DNS requests go direct while application traffic uses another route, the resulting answer may not match the path you expected. Consistency is more important than selecting a fashionable resolver.
- Export or duplicate your current Clash profile before changing DNS, TUN, or rule order.
- Enable only one capture method, then confirm that the Clash dashboard shows connections from the browser and the relevant desktop helpers.
- Apply the collaboration rules and check that local sites, private addresses, and approved intranet domains still use
DIRECT. - Restart the affected application after changing TUN or DNS so old sockets do not conceal the new behavior.
- Save a dated backup after the workflow passes all tests, and note which policy group was selected during the test.
Test, troubleshoot, and maintain the setup
Test each product with a small repeatable checklist rather than judging the entire system by one page load. In Notion, edit a block, open a database view, upload an attachment, and confirm that the page remains usable after a reconnect. In Figma, open a shared file, pan across a complex canvas, check fonts and images, and perform a small export. In Miro, create or move an object, wait for synchronization, reload the board, and test a media or document preview. Record whether the failure is authentication, live editing, asset delivery, or export; each category may use a different hostname.
Use the Clash connection log as a timeline. Filter by the service name, then compare the hostname with the rule that matched it and the policy group that handled it. If a request is unexpectedly DIRECT, check rule order and whether the domain is an alias or a CDN hostname. If it reaches the correct group but times out, test another node inside that group and compare latency, packet loss, and sustained download speed. If changing nodes fixes only large assets, keep the application rule stable and consider a separate asset policy rather than rotating the entire workflow.
Common failure patterns
- Login works, editing fails: inspect websocket or long-lived connections, then verify that TUN captures the desktop helper or browser process.
- Workspace loads, images are missing: identify the image or object-storage hostname in the log and avoid assuming it belongs to the main service domain.
- Figma canvas is visible but fonts are absent: check font delivery domains, local font helper permissions, and whether a browser extension is intercepting requests.
- Miro updates arrive late: compare the selected route’s latency and stability with a manual alternative; automatic health checks may not represent realtime traffic quality.
- Local websites stop opening: review private-network, LAN, intranet, and local DNS rules before changing proxy nodes.
- Everything works until the laptop wakes: restart the Clash service or reconnect TUN, then inspect stale DNS and half-open sockets created before sleep.
When a rule becomes reliable, document why it exists. Write a short comment in your configuration or maintain a separate note containing the hostname, observed function, date tested, and preferred group. This prevents future cleanup from deleting a rule that looks redundant. Recheck the workflow after a major Clash or Mihomo update, a provider profile refresh, or a change in corporate DNS. A profile that works in March may classify a CDN differently in August.
Some competing clients and traditional VPN tools make this workflow harder by hiding domain-level decisions behind one global switch, offering limited websocket diagnostics, or treating DNS and tunnel capture as unrelated toggles. Others provide a polished one-click connection but become difficult to tune when only Figma assets fail or when Miro needs a different route from local work websites. Clash offers a more transparent alternative: policy groups, ordered domain rules, observable connection logs, configurable DNS modes, and Mihomo TUN support let you isolate Notion, Figma, and Miro traffic without forcing every local service through the same exit. If you want to build this workflow with a supported client, visit the download page and choose the platform that matches your device.
→ Download Clash for free and enjoy a better internet experience