By Ira Bodnar, Ryze AI · Updated · Connector tested November 13, 2026, address only
Quick answer
The main way to connect OpenClaw to Thrad is the Ryze AI connector. Create a Thrad API key, paste it into the Thrad card in Ryze AI, then save https://connector.get-ryze.ai/mcp as a remote server in OpenClaw (openclaw mcp set, or an entry in openclaw.json) and sign in with openclaw mcp login ryze.
- In the Thrad Platform dashboard create an API key under Settings > API keys
- In Ryze AI open Integrations, click Connect on the Thrad card and paste the key
- In OpenClaw save the server named ryze (CLI or openclaw.json) and run openclaw mcp login ryze
- Start a new OpenClaw turn and ask about your Thrad campaigns, ads or insights
OpenClaw to Thrad at a glance
| Fact | Detail |
|---|---|
| Main way to connect | The Ryze AI connector: one hosted MCP address, saved in OpenClaw as a remote server with transport streamable-http |
| Server address | https://connector.get-ryze.ai/mcp |
| Sign-in | OAuth in the browser through openclaw mcp login. A Ryze AI API key in an Authorization header also works. Thrad itself is connected with an ak_ API key in Ryze AI |
| Thrad tools | 5: two read, two write, one delete |
| Can it change Thrad | Yes. The write tool creates and updates campaigns, ad groups and ads and activates or pauses them. A separate delete tool archives them |
| Other data in the same chat | Thrad insights combine Thrad supply and ChatGPT (OpenAI Ads) supply. Tools for other platforms connected in the same Ryze AI workspace use the same address |
| OpenClaw plan needed | None. OpenClaw is open source (MIT) and self-hosted |
| Ryze AI price | Paid Ads is $89 per month with a 7-day free trial. Adding the connector is free |
| Thrad's own option for OpenClaw | None for an ad account. Thrad publishes a documentation MCP server (https://docs.thrad.ai/mcp) that searches its public docs only |
| Where Thrad works in Ryze AI | In the Ryze AI chat and through the connector. The help center lists Thrad as chat only, so it does not appear on the Ads Performance dashboard or in Managed Campaigns |
| Last checked | Sources read November 14 and 15, 2026. The steps inside OpenClaw were not run |
Before you start
- An OpenClaw Gateway you control. OpenClaw is a self-hosted open-source gateway under the MIT license, with no plan tier. The CLI steps edit the state of the Gateway on your machine or server. Adding a server from the chat composer or signing in from Settings needs administrator access.
- A Thrad account and an owner or admin role. The connection uses a Thrad API key (it starts with ak_). Thrad's documentation says only organization owners and admins can create, rotate or revoke a key, and that an organization has one active key.
- A Ryze AI workspace on a plan or free trial. Paid Ads is $89 per month with a 7-day free trial. Adding the connector is free, but the Thrad tools need an active plan or trial.
- A card on the Thrad organization, if you plan to create campaigns. Thrad's documentation says creating a campaign needs a default card and charges a $10 deposit. That is a Thrad charge and not a Ryze AI price.
How to connect OpenClaw to Thrad
Two steps happen in Thrad and Ryze AI, two in OpenClaw, and the last is your first prompt. The connector URL is the same for every user.
Create a Thrad API key
In the Thrad Platform dashboard go to Settings > API keys, create a key and copy it. It starts with ak_. Thrad shows the full secret only once, and an organization has one active key.
Connect Thrad in Ryze AI
Create a Ryze AI account or sign in, open Integrations, click Connect on the Thrad card, paste the key into API key and click Continue. Ryze AI verifies the key and the card shows Connected.
Save the Ryze AI server in OpenClaw
Stop the Gateway if it is running (openclaw gateway stop), then run the openclaw mcp set command shown below. Or add the same entry under mcp.servers in ~/.openclaw/openclaw.json, which the Gateway watches and applies.
openclaw gateway stop openclaw mcp set ryze '{"url":"https://connector.get-ryze.ai/mcp","transport":"streamable-http","auth":"oauth"}' openclaw mcp login ryze openclaw gateway startSign in from OpenClaw
Run openclaw mcp login ryze, approve the Ryze AI sign-in in the browser and click Allow connection. If the browser is on another machine, pass the returned code back with --code. Then start the Gateway again.
Ask your first question
Start a new turn in OpenClaw and try: Show my Thrad ad account and list every campaign with its status and review status.
The Ryze AI help center page for Thrad covers the Thrad side of the connection. OpenClaw does not have its own Ryze AI help page, and Ryze AI was not tested there. The Ryze AI developers page covers API keys.
Sources: OpenClaw MCP registry commands, OpenClaw MCP transports, Thrad API authentication, Ryze AI Thrad guide. Checked November 14 and 15, 2026.
CLI, openclaw.json and Settings
The CLI
The openclaw mcp set and openclaw mcp login commands refuse to run while a Gateway owns the state directory, a rule OpenClaw added in release 2026.10.10 on November 3, 2026. The login prints a browser address. Approve the Ryze AI sign-in there and click Allow connection. If the browser runs on another machine or cannot reach the printed loopback address, copy the returned code and pass it back with --code.
openclaw gateway stop
openclaw mcp set ryze '{"url":"https://connector.get-ryze.ai/mcp","transport":"streamable-http","auth":"oauth"}'
openclaw mcp login ryze
openclaw gateway startopenclaw.json
Merge this into ~/.openclaw/openclaw.json, which is JSON5, so comments and trailing commas are fine. OpenClaw's documentation says the Gateway watches the file and applies the change, and that unknown keys make the Gateway refuse to start. Run openclaw mcp login ryze afterwards. When transport is left out, OpenClaw uses sse, so set it.
{
mcp: {
servers: {
ryze: {
url: "https://connector.get-ryze.ai/mcp",
transport: "streamable-http",
auth: "oauth",
},
},
},
}openclaw.json with a Ryze AI API key
If you prefer a Ryze AI API key to the browser sign-in, drop auth: "oauth" and add a headers entry that reads the key from an environment variable. OpenClaw ignores a static Authorization header while auth: "oauth" is on, so use one or the other. The Ryze AI developers page covers API keys. Keep the key out of files you commit.
{
mcp: {
servers: {
ryze: {
url: "https://connector.get-ryze.ai/mcp",
transport: "streamable-http",
headers: { Authorization: "Bearer ${RYZE_API_KEY}" },
},
},
},
}Settings and the chat composer
A third route works while the Gateway runs. In the Control UI open Settings > MCP, select Add server under Configured servers, give it the name ryze, pick Streamable HTTP and enter the URL. OAuth metadata and tool filters go in the scoped config editor lower on the same page. An administrator can then select Sign in for the server. From a Control UI chat you can also use + > Plugins > Add MCP server, with the scope This session or Everywhere. OpenClaw's documentation says Sign in from Settings works only when the Control UI is opened on the Gateway's own HTTPS address, its loopback address or its Tailscale address.
Check the saved server
Saving a definition proves nothing about reachability. OpenClaw's documentation points to openclaw mcp doctor ryze --probe, which opens a live connection and lists the tools. openclaw mcp tools ryze lists them and takes --include and --exclude patterns.
OpenClaw and MCP: the facts in one table
| Topic | What OpenClaw's documentation says |
|---|---|
| Where servers are added | Control UI Settings > MCP > Add server, the chat composer (+ > Plugins > Add MCP server), the CLI (openclaw mcp add or set), or mcp.servers in ~/.openclaw/openclaw.json |
| Transports | stdio, SSE and Streamable HTTP. A URL entry without transport uses sse |
| Sign-in | auth: "oauth" on an HTTP server, then openclaw mcp login <name> or Sign in on Settings > MCP. Credentials are shared and operator-managed by default |
| Per-person sign-in | oauth.identity: "per-requester" gives each sender their own account and needs gateway.publicOrigin set to an HTTPS origin |
| Before sign-in | OpenClaw leaves only that server out of the agent's tools. The turn does not fail |
| Tool policy | MCP tools follow OpenClaw's tool profiles. The coding and messaging profiles include them, minimal hides them, and tools.deny: ["bundle-mcp"] turns them off. + > Plugins > Tool access denies single tools for one session |
| Tool filter | toolFilter.include keeps only matching tool names and toolFilter.exclude hides matching names. Exact names or simple * globs |
| Admin controls | Adding a server from the composer and signing in from Settings need administrator access. enabled: false keeps a definition but excludes it |
| Tool names | Tools reach the agent as <server>__<tool>, for example docs__search |
| Plan | None. Open source under the MIT license, self-hosted |
Sources: OpenClaw MCP documentation, OpenClaw config extensions, OpenClaw configuration, OpenClaw tool policy. Checked November 15, 2026.
Claude and other guides
The same URL works in other MCP clients. Thrad is connected once in Ryze AI, so nothing changes on the Thrad side.
Claude
Add https://connector.get-ryze.ai/mcp as a custom connector under Customize > Connectors and start a new chat. If the Thrad tools do not appear, remove the connector, add it again and start a new chat. Thrad is connected once in Ryze AI, so nothing changes on the Thrad side.
Other guides for Thrad
Claude Code, ChatGPT, Cursor and Codex each have their own Thrad guide, linked under Related guides below. They use the same URL and the same 5 tools.
Thrad key and access
Two sign-ins are involved: Thrad for the account and Ryze AI for the connector. Thrad publishes no scope list for its Advertiser API and describes one key type, the ak_ key, so what OpenClaw can reach is decided by that key and by the approval and tool settings you choose.
- Who can create the key
- Only Thrad organization owners and admins can create, rotate or revoke a key. Each request carries the access of exactly one organization, and an organization has one active key.
- What the key covers
- Reading the ad account, the asset catalog and insights at ad account, campaign, ad group and ad level, and creating, updating, activating, pausing and archiving campaigns, ad groups and ads. Thrad's documentation describes no read-only API key.
- Who shares the sign-in
- OpenClaw stores the Ryze AI sign-in as shared, operator-managed credentials by default. Senders on that Gateway share that one sign-in, so they share the Thrad tools it exposes, write and delete included. Per-person sign-in needs oauth.identity: "per-requester" and gateway.publicOrigin.
- How to limit changes
- In OpenClaw, use toolFilter.exclude or + > Plugins > Tool access to hide the write and delete tools. In the Ryze AI chat, the Tool approvals setting starts on Skip all approvals, and you can switch it to Approve write & delete or Manual approve. In Claude, set write and delete tools to Needs approval under Customize > Connectors > Ryze AI.
What the key covers
Thrad's documentation names these areas for an API key. The table says what each one covers and whether the Ryze AI Thrad tools reach it.
| Area | What Thrad documents | Ryze AI tool |
|---|---|---|
| Ad account | Read only. One default ad account per organization | thrads__read |
| Campaigns, ad groups and ads | Create, update, activate, pause and archive | thrads__read, thrads__write, thrads__delete |
| Asset catalog | Read only over the API. Ads point at an existing asset | thrads__read |
| Insights | Impressions, clicks, spend, CTR, CPC and CPM at ad account, campaign, ad group and ad level | thrads__insights |
What is not included
The key does not cover several things. Thrad's documentation lists them, and so does the cannot list of the Ryze AI Thrad connection.
| Area | What Thrad documents | With Ryze AI |
|---|---|---|
| Key management | Creating, rotating and revoking keys needs a dashboard session and sits outside the /v1/ API | Not reachable |
| Assets and image upload | Assets are read-only over /v1/ and the API accepts no image bytes and no image URL. thrads__uploadFromUrl currently gets a 404 from Thrad | Use an existing asset |
| Approving a campaign | Thrad reviews new campaigns. Until approval, activate returns the code campaign_not_approved | Thrad's review |
| Billing setup | The default card is added in the Thrad Platform dashboard | Dashboard only |
| Targeting below country level | Region, state and DMA targeting are not supported | Country level only |
Managed Campaigns and the Ads Performance dashboard in Ryze AI are not available for Thrad, because the Ryze AI help center lists Thrad as chat only. Archiving a campaign cannot be undone, while an archived ad group or ad reads back as paused and can be activated again.
Sources: Thrad API authentication, Thrad assets, Thrad targeting, Ryze AI Thrad guide. Checked November 14, 2026.
Removing access
To stop OpenClaw from reaching Thrad, run openclaw mcp logout ryze to clear the stored tokens, or openclaw mcp unset ryze to remove the server, or set enabled: false on the entry in openclaw.json. To cut Ryze AI's access to Thrad altogether, revoke the key in the Thrad Platform dashboard. Thrad's documentation says revoking disables a key immediately with no grace period. Creating a new key makes the old one stop working, and the Thrad card in Ryze AI then needs Reconnect.
What OpenClaw can do in Thrad
The Ryze AI connector exposes five Thrad tools: two read, two write and one delete. Writes and deletes are separate tools, so you can hide or approve them separately.
| Tool | Type | What it does |
|---|---|---|
| thrads__read | Read | Reads any Thrad resource, such as the ad account, campaigns, ad groups, ads and assets. |
| thrads__insights | Read | Returns aggregated insights that combine Thrad and ChatGPT (OpenAI Ads) supply: impressions, clicks, spend, CTR, CPC and CPM. |
| thrads__write | Write | Creates or updates Thrad campaigns, ad groups and ads, and activates or pauses them. New campaigns are submitted for Thrad's review before they go live. |
| thrads__uploadFromUrl | Write | Uploads an image from a URL to get a file id for an ad creative. Thrad currently answers this call with a 404, so existing assets are used instead. |
| thrads__delete | Delete | Archives a Thrad campaign, ad group or ad. Archiving a campaign is irreversible. |
Some limits apply to every client, OpenClaw included. Thrad targeting is country level only, images cannot be uploaded reliably, and an archived campaign cannot be brought back. Every thrads tool takes an optional connection_id, which is how several connected Thrad accounts are told apart. There is no default accounts picker for Thrad, so name the account in your prompt.
OpenClaw's documentation does not say that the built-in runtime asks before each MCP tool call. MCP tools follow the tool profiles and tool policy, so hide the write and delete tools with toolFilter.exclude, or deny single tools for a session under + > Plugins > Tool access, if you want OpenClaw to read only. Which name form the filter matches was not tested, so check the result with openclaw mcp tools ryze. Separately, the Ryze AI chat has its own Tool approvals setting that starts on Skip all approvals.
Limits in numbers
| Limit | Number |
|---|---|
| Thrad requests | 1,000 per hour per key across all /v1/ endpoints. The count is per key, not per IP and not per organization. |
| Campaign creation | 60 per hour per key, on top of the 1,000 per hour bucket. |
| Over the limit | HTTP 429 with the code rate_limit_exceeded and the message Request was throttled. |
| List size | 1 to 500 rows on campaign, ad group and ad lists, default 20. Ad groups are listed one campaign at a time and ads one ad group at a time. |
| Insights window | The trailing 30 days by default. One window spans at most 366 days, or 25 hours with hourly granularity. Rows page by limit, 1 to 2,000, default 20. |
| ChatGPT supply | ChatGPT (OpenAI Ads) insights are daily only. Asking for them with hourly granularity is an error. |
| Insights metrics | 6 fields: impressions, clicks, spend, CPC, CPM and CTR. Money is counted in micros, where $1 is 1,000,000. |
| Budget and bid | A campaign budget starts at $0.99 with no upper cap. Max bid per ad group runs from $0.0001 to $100, and a per-country bid floor can still reject a bid in that range. |
| Campaign deposit | Thrad charges a $10 deposit on the organization's default card when a campaign is created. |
| OpenClaw request timeout | 60 seconds per request by default. Tool listing at session start uses 10 seconds when requestTimeoutMs is not set. |
| OpenClaw runtimes | At most 256 OpenClaw-managed MCP runtimes per Gateway and 448 MCP child processes. No limit on the number of servers or tools is stated. |
| Server instructions | OpenClaw passes at most the first 2,000 characters of a server's instructions to the embedded agent. |
Sources: Thrad API authentication, Thrad insights, Thrad campaigns, Thrad ad groups, OpenClaw config extensions. Checked November 14 and 15, 2026.
First 3 prompts to run
1. Review the account and its campaigns
Show my Thrad ad account with its name, currency and timezone, then list every campaign with its status and review status.
Tools used: thrads__read
What comes back: The account details and one line per campaign. Review status is Thrad's own verdict (in_review, approved or rejected), separate from the status you control. Nothing is changed.
2. Report delivery by supply
Report impressions, clicks, spend, CTR, CPC and CPM for the last 30 days, and split the numbers between Thrad supply and ChatGPT (OpenAI Ads) supply.
Tools used: thrads__insights
What comes back: Rows for the six metrics, with a Thrad row and an OpenAI row that sum to the combined total when you ask for the split. Nothing is changed.
3. Pause an ad group
Pause the ad group named in my message, then read it back and tell me its status.
Tools used: thrads__read, thrads__write
What comes back: A write call that pauses the ad group, then a read of its status so you can compare it with what you asked for. Name the Thrad account in the prompt if several are connected.
This page shows no sample output. The packs behind it record no call to a Thrad account, so each prompt describes what comes back and the figures are for you to read from your own account. Replace the names in the prompts with your own.
10 jobs to hand to OpenClaw
Ten jobs to hand to your OpenClaw agent, each matched to the tool it uses and whether the default connection covers it. None of them was run for this page.
| Job | Ask OpenClaw | Tool | Works on the default connection |
|---|---|---|---|
| Account check | Show my Thrad ad account: name, currency and timezone. | thrads__read | Yes |
| Campaign list | List every campaign with its status and review status. | thrads__read | Yes |
| Ad group and ad list | List the ad groups and ads under a campaign, with context hints and max bid. | thrads__read | Yes |
| Asset list | List the assets already in the account that an ad can use. | thrads__read | Yes |
| Delivery report | Report impressions, clicks, spend, CTR, CPC and CPM for the last 30 days. | thrads__insights | Yes |
| Supply comparison | Compare Thrad supply with ChatGPT (OpenAI Ads) supply by splitting insights by source. | thrads__insights | Yes |
| New campaign | Create a campaign with an ad group and a card or poll ad that uses an existing asset. | thrads__write | Yes. Thrad reviews the campaign before it goes live |
| Edit | Update a campaign, ad group or ad. | thrads__write | Yes |
| Pause or activate | Pause or activate a campaign, ad group or ad. | thrads__write | Yes. Activate only resumes a paused, approved campaign |
| Archive | Archive a campaign, ad group or ad. | thrads__delete | Yes. Archiving a campaign is irreversible |
3 ways to connect OpenClaw to Thrad
Thrad publishes no MCP server, connector or plugin that reaches an advertiser account. This guide uses the Ryze AI connector as the main way: it needs no code and no Thrad key handling beyond one paste. Thrad's REST API is the other way to reach an account, and its documentation MCP server is a companion that reaches no account. The checks were made on November 14 and 15, 2026, and none of the three was run inside OpenClaw.
| Way | Runs where | Setup in OpenClaw | What it reaches in Thrad | Price |
|---|---|---|---|---|
| Ryze AI connector (the main way) | Hosted. One address for every user | mcp.servers entry with url, transport streamable-http and auth oauth, then openclaw mcp login ryze | Ad account, campaigns, ad groups, ads and assets to read, insights, creating and updating, and archiving. 5 tools | Paid Ads is $89 per month with a 7-day free trial |
| Thrad Advertiser API from your own code | Thrad's REST API at https://api.thrad.ai/v1 | No Thrad connector exists. Thrad describes the API as called from your own code with an ak_ key sent as a Bearer token. You write and run the calls yourself | The same account areas, with the same limits: 1,000 requests per hour per key and 60 campaign creations per hour | No API fee stated. Thrad charges a $10 deposit per campaign created |
| Thrad Docs MCP server | Hosted by Thrad's docs host | Save https://docs.thrad.ai/mcp as a remote server. Thrad names Cursor for install and no sign-in is needed. It was not tried in OpenClaw | Public documentation only. 3 tools: search_thrad_docs, query_docs_filesystem_thrad_docs and submit_feedback. No campaigns, ads or spend | No price stated |
Options that do not fit this job
- Composio, Zapier MCP and the reporting connectors. Composio's toolkit list, Zapier's app directory, Windsor.ai, Coupler.io, Supermetrics and Porter Metrics had no Thrad entry on November 14, 2026.
- An open-source Thrad MCP server. A GitHub search for thrad mcp returned no repositories on November 14, 2026, and the npm, PyPI and official MCP Registry searches found no Thrad server.
- ClawHub. OpenClaw's public registry for skills and plugins, with its own Add MCP integration option for administrators. A ClawHub search on November 15, 2026 found no Ryze AI plugin, so the connector is added by URL.
- Thrad Agent Skills. The thrad-card, thrad-poll and thrad-carousel skills are for publishers who put Thrad ads into their own chat app. They do not manage advertiser campaigns.
Which one to pick
Ryze AI is the best way to connect OpenClaw to Thrad for reading and managing an ad account from chat, with one caveat: it cannot upload images, change targeting below country level or bring back an archived campaign, and Thrad is chat only in Ryze AI. Use Thrad's API directly when you want your own code, scheduling or a read-only flow, since Thrad describes no read-only key. Add the Thrad Docs server when you want OpenClaw to answer questions about Thrad's API from the documentation.
Sources: Thrad Advertiser API overview, Thrad Docs MCP server, Thrad Agent Skills, ClawHub, OpenClaw MCP documentation, Composio, Zapier MCP, Ryze AI Thrad guide. Checked November 14 and 15, 2026.
OpenClaw and Thrad: dates that matter
| Date | What happened |
|---|---|
| March 31, 2026 | OpenClaw release 2026.3.31 added remote MCP servers by URL under mcp.servers, with auth headers and an optional streamable-http transport. |
| April 23, 2026 | ExchangeWire PressBox published Thrad Is Now Live as an Advertiser Inside ChatGPT. The page does not say who wrote it, so it is press coverage and not a Thrad changelog. |
| April 27, 2026 | OpenClaw release 2026.4.25 made changes under mcp.* apply without a restart. |
| August 7, 2026 | Thrad's docs sitemap lists its Advertiser API pages (overview, quickstart, authentication, campaigns, insights, assets) as last modified. This is a last-modified date, not a launch date. |
| August 31, 2026 | OpenClaw 2.0 (release 2026.8.1) added servers from composer connectors and per-requester OAuth for shared MCP servers. |
| September 23, 2026 | OpenClaw release 2026.9.6 let administrators sign in to MCP connectors directly from Settings. |
| November 3, 2026 | OpenClaw release 2026.10.10 protects MCP configuration changes from conflicting writes while the Gateway runs. Commands without a live path need the Gateway stopped first. |
| November 14, 2026 | Thrad's Docs MCP server answered at https://docs.thrad.ai/mcp as version 1.0.0 with 3 tools. This is the date of the check, not a launch date. Thrad has no changelog page and publishes no launch date for its Advertiser API or an account-level MCP server. |
Sources: OpenClaw releases, OpenClaw changelog 2026.3.31, OpenClaw release 2026.10.10, Thrad docs sitemap, ExchangeWire, April 23, 2026. Checked November 14 and 15, 2026.
How this guide was tested
This guide rests on documentation and endpoint checks, not on a run inside OpenClaw. The table lists what was read or run and what came back.
| What was run | What came back |
|---|---|
| OpenClaw's MCP documentation pages and release notes | Read on November 15, 2026. The surfaces, transports, sign-in rules, limits and release dates on this page come from them. |
| Thrad's Advertiser API documentation | Read on November 14, 2026. The key rules, limits and error codes on this page come from it. |
| The connector address, without signing in | A request to https://connector.get-ryze.ai/mcp came back as HTTP 401, recorded in the OpenClaw fact pack as a connector test of November 13, 2026. OpenClaw's documentation says a 401 on a Streamable HTTP server is reported as authentication refused (401), but that pairing was not observed. |
| Thrad's Docs MCP server, without signing in | On November 13 and again on November 14, 2026 the server named Thrad Docs reported version 1.0.0, protocol version 2025-06-18 and 3 tools. |
| ClawHub search for Ryze AI | No Ryze AI plugin on November 15, 2026. |
Not run for this page: every step inside OpenClaw, because no Gateway was installed, no server was saved and no sign-in was attempted. Also not run: any call to a Thrad account and any Ryze AI Thrad tool, because no Thrad account was connected. The tool names come from the Ryze AI app source and were not seen live. The steps follow OpenClaw's documentation as it read on November 15, 2026.
OpenClaw and Thrad errors and fixes
Error messages from the vendors' documentation
These messages are copied from Thrad's and OpenClaw's documentation. They were not seen in a test of this guide, and Thrad's messages come from the API, so Ryze AI may word them differently.
| Message or code | Source | What it means |
|---|---|---|
| authentication refused (401) | OpenClaw docs | A Streamable HTTP server refused initialization with HTTP 401. Run openclaw mcp login ryze. |
| authorization-required | OpenClaw docs | openclaw mcp status can report this. Run openclaw mcp login ryze again. |
| insufficient_scope | OpenClaw docs | The server rejected a token for lack of scope. OpenClaw asks for openclaw mcp login ryze instead of refreshing. |
| Invalid API key. | Thrad docs, 401 invalid_api_key | The Thrad key is malformed, unknown, expired or revoked. Create a key in Thrad and use Reconnect on the Thrad card. |
| Request was throttled. | Thrad docs, 429 rate_limit_exceeded | More than 1,000 requests per hour on the key, or more than 60 campaign creations per hour. |
| Set up a default card in billing settings before creating a campaign. | Thrad docs, 400 billing_not_configured | The organization has no default card. Nothing is created and there is no charge. |
| Only an approved, paused campaign can be activated. | Thrad docs, 400 campaign_not_approved | The campaign is still in review or was rejected. Thrad does the review. |
Other problems
- The Thrad tools do not appear in OpenClaw
- Until the sign-in exists, OpenClaw leaves the server out of the agent's tools and the turn does not fail. Run openclaw mcp login ryze, then openclaw mcp doctor ryze --probe. If Thrad was connected in Ryze AI after the server was saved, remove the entry with openclaw mcp unset ryze, save it again, sign in again and start a new turn.
- openclaw mcp set or login refuses to run
- A Gateway owns the state directory. Run openclaw gateway stop first and start it again afterwards, or use Settings > MCP in the Control UI, which works while the Gateway runs.
- The server saved but nothing connects
- Check that transport is streamable-http. A URL entry without transport uses sse, and OpenClaw's documentation says a malformed config can stop the Gateway from starting.
- Sign in from Settings does not finish
- OpenClaw's documentation says Sign in works only when the Control UI is opened on the Gateway's own HTTPS address, its loopback address or its Tailscale address. A separately hosted UI cannot receive the callback.
- My Ryze AI API key header is ignored
- OpenClaw ignores a static Authorization header while auth: "oauth" is on. Drop auth: "oauth" when you use a key.
- Thrad stopped working after I made a new key
- Thrad allows one active key per organization, and the old key stops working when a new one is created. Open the Thrad card in Ryze AI and use Reconnect with the new key.
- A campaign will not activate
- Thrad reviews new campaigns. Until the review approves the campaign, activate returns campaign_not_approved.
- Creating a campaign fails on billing
- Thrad's documentation says a campaign needs a default card on the Thrad organization and charges a $10 deposit. Add the card in the Thrad Platform dashboard.
- An image upload fails
- Thrad currently answers the upload call with a 404 and assets are read-only over its API. Ask the agent to use an asset that already exists.
- A request times out
- OpenClaw's request timeout is 60 seconds by default. For Thrad insights, Thrad documents a timeout code clickhouse_timeout when the query is too wide, so narrow the scope, window or granularity.
- The tools reply that a subscription is required
- The Thrad tools need an active Ryze AI plan or free trial. Start the trial, then retry.
Terms on this page
- MCP
- The Model Context Protocol, an open standard for connecting AI applications to outside systems. OpenClaw supports it for servers it saves under mcp.servers.
- MCP server
- A program that offers tools over MCP. The Ryze AI connector is an MCP server that runs on Ryze AI's side, so there is nothing to install.
- OpenClaw Gateway
- The single process of OpenClaw that runs on your own machine or server, connects chat apps to AI agents and holds the saved MCP servers.
- Streamable HTTP
- One of the three MCP transports OpenClaw supports, alongside stdio and SSE. Set transport to streamable-http for the Ryze AI address.
- Supply (source)
- Where a Thrad ad was shown: Thrad's own network or ChatGPT through OpenAI Ads. Insights combine both unless you ask for one.
- Context hints
- Short phrases on a Thrad ad group that steer where its ads serve. Each ad group needs at least one.
- Review status
- Thrad's moderation verdict on a campaign: in_review, approved or rejected. It is separate from the status you control.
Source for MCP: modelcontextprotocol.io. Checked November 16, 2026.
Frequently asked questions
How do I connect OpenClaw to Thrad?
Create a Thrad API key under Settings > API keys, paste it into the Thrad card in Ryze AI, then save https://connector.get-ryze.ai/mcp in OpenClaw as a remote server with transport streamable-http and auth oauth, and run openclaw mcp login ryze.
Is there a Thrad connector for OpenClaw?
Yes, the Ryze AI connector at https://connector.get-ryze.ai/mcp has 5 Thrad tools that OpenClaw can use as a remote MCP server. Thrad itself publishes no connector, plugin or MCP server for an ad account, and a ClawHub search on November 15, 2026 found no Ryze AI plugin, so you add it by URL.
Is there a Thrad MCP server for OpenClaw?
The Ryze AI connector is the MCP server that reaches a Thrad ad account. Thrad also hosts a documentation MCP server at https://docs.thrad.ai/mcp with 3 tools, which searches its public docs and reaches no campaigns or spend. It was not tried in OpenClaw.
What is the Thrad MCP server URL for OpenClaw?
For the Ryze AI connector the address is https://connector.get-ryze.ai/mcp, the same for every user. Thrad's documentation server is https://docs.thrad.ai/mcp and needs no sign-in. Thrad's REST API is https://api.thrad.ai/v1, which is not an MCP address.
Does OpenClaw support remote MCP servers?
Yes. OpenClaw saves a remote server by URL under mcp.servers with transport streamable-http or sse, from the Control UI, the chat composer, the CLI or ~/.openclaw/openclaw.json. Release 2026.3.31 on March 31, 2026 added remote servers by URL.
Which OpenClaw plan do I need?
None. OpenClaw is open source under the MIT license and self-hosted, with no plan tier. Adding a server from the chat composer and signing in from Settings need administrator access on the Gateway.
Can OpenClaw change my Thrad campaigns?
Yes. Through Ryze AI, OpenClaw can create and update campaigns, ad groups and ads and activate or pause them with the write tool. New campaigns go to Thrad for review before they go live. Archiving goes through the separate thrads__delete tool, and archiving a campaign cannot be undone.
Can I make OpenClaw read only in Thrad?
Yes, on the OpenClaw side. Hide the write and delete tools with toolFilter.exclude or deny them for a session under + > Plugins > Tool access. Thrad's documentation describes no read-only API key, so the limit comes from OpenClaw's tool settings and not from Thrad.
Does OpenClaw ask before it runs a Thrad write tool?
OpenClaw's documentation does not say that its built-in runtime asks before each MCP tool call, and per-call prompts are documented for Codex runs only. Use tool filters if you want to stop changes. The Ryze AI chat has its own Tool approvals setting that starts on Skip all approvals.
Can several people on my OpenClaw Gateway use my Thrad account?
By default OpenClaw stores the Ryze AI sign-in as shared, operator-managed credentials, so senders on that Gateway share it. Per-person sign-in needs oauth.identity: "per-requester" and gateway.publicOrigin set to an HTTPS origin.
How much does it cost to connect OpenClaw to Thrad?
Adding the connector costs nothing, but the Thrad tools need an active Ryze AI plan or free trial. Paid Ads is $89 per month with a 7-day free trial, SEO Starter is $129 per month with a 3-day free trial, Agency is $279 per month and Shopify CRO is $1,499 per month. Annual billing takes 30% off the first year. Thrad charges its own $10 deposit when a campaign is created.
Why do the Thrad tools not show up in OpenClaw?
Until the sign-in exists, OpenClaw leaves the server out of the agent's tools. Run openclaw mcp login ryze and openclaw mcp doctor ryze --probe. If Thrad was connected in Ryze AI after you saved the server, remove the entry, save it again and start a new turn.
Why did the Thrad connection stop working?
Thrad allows one active key per organization, and the old key stops working when a new one is created. Use Reconnect on the Thrad card in Ryze AI with the new key.
Can I connect more than one Thrad account?
Yes. Several Thrad accounts can be connected to one workspace with View > Add one more account on the Thrad card, and every thrads tool takes an optional connection_id. There is no default accounts picker for Thrad, so name the account in your prompt.
How do I disconnect Thrad from OpenClaw?
Run openclaw mcp logout ryze to clear the tokens, openclaw mcp unset ryze to remove the server, or set enabled: false on the entry. To remove Thrad access altogether, revoke the key in the Thrad Platform dashboard.