Trend Seeker Now Works Inside Claude Code and Codex
Use Claude Code or Codex to find and validate business ideas with Trend Seeker's dated market evidence through a read-only MCP? server.
5 min read
Introduction
Claude Code and Codex can now use Trend Seeker to find business ideas, inspect dated market evidence, validate a concept, and compare a shortlist. The new public plugin connects both agents to Trend Seeker's hosted, read-only MCP server.
This is useful when you want an AI agent to do more than brainstorm plausible products. The agent can search Trend Seeker's opportunity database, open the research summary for a result, and retrieve the stored sources behind it. You still have to judge the evidence and talk to customers. The plugin gives the agent a better starting point for that work.
Three Business-Idea Research Workflows
The first release packages three focused workflows. They use the same tools in Claude Code and Codex, so the research method stays consistent across both agents.
| Workflow | Question it answers | What the agent checks |
|---|---|---|
| Find opportunities | What could I build for this market, audience, or problem? | Matching ideas, research summaries, scores when available, and supporting sources. |
| Validate an idea | Does Trend Seeker contain direct or adjacent evidence for my concept? | Closest opportunities, dated evidence, mismatches, and the largest unanswered assumptions. |
| Compare opportunities | Which option best fits my constraints? | The same available fields for two to five ideas, followed by evidence and risks for each. |
The workflows deliberately avoid invented market sizes, revenue estimates, and success probabilities. Missing evidence stays missing. A weak match does not become validation just because an agent can write a confident explanation.
How the Agent Gets Current Market Evidence
The plugin combines reusable research instructions with six MCP tools. The instructions tell the agent how to search, inspect finalists, preserve source dates and links, and explain gaps. The hosted server supplies the structured Trend Seeker data.
flowchart LR A[Your research question] --> B[Claude Code or Codex] B --> C[Trend Seeker research skill] C --> D[Read-only MCP tools] D --> E[Ideas and dated evidence] E --> B
The agent chooses a research workflow, calls the relevant read-only tools, and turns the returned ideas and sources into an answer you can inspect. The MCP server cannot modify Trend Seeker data.
MCP defines tools as model-controlled functions exposed by a server. In this case, those functions search and retrieve a bounded research dataset. They do not browse arbitrary websites while answering a request.
A Real Search from the Public MCP Server
On September 21, 2026, I asked the public endpoint to find “developer tools for teams adopting AI coding agents.” It returned three matching opportunities, led by AI-Native Engineering Workflow Implementation Service. The other results covered AI-agent token cost optimization and engineering-workflow adoption.
I then asked for three supporting records behind the leading result. Trend Seeker returned job-ad evidence observed on September 19 and 20. The records described adopting AI-assisted development tools across an engineering organization, adding generative AI to developer workflows, and setting standards for reviewing AI-written code. Their original publication dates were unavailable, so the agent should preserve that uncertainty instead of presenting the observation dates as publication dates.
This example shows the practical difference between asking an AI to generate business ideas and asking it to research them. The first produces possibilities. The second connects a possibility to records you can open, challenge, and follow up on.
How to Use Trend Seeker in Claude Code
The Trend Seeker plugin repository is public. Clone it, enter the directory, and start Claude Code with claude --plugin-dir .. The bundled MCP connection uses anonymous read-only access, so the initial workflow does not require a Trend Seeker account or API? key.
You can then ask Claude Code to find opportunities for a market, validate a defined idea, or compare the strongest results for your constraints. The repository includes exact slash-command examples and troubleshooting steps. Anthropic also documents how Claude products connect to MCP servers.
How to Use Trend Seeker in Codex
The same package includes a Codex plugin manifest, the shared research skills, and the hosted MCP configuration. Add the public repository as the source of a configured Codex marketplace entry, install the plugin, and start a new session before using its tools. OpenAI's plugin documentation explains how skills and MCP servers work together in Codex.
The public OpenAI directory listing is still in draft as of September 21, 2026. The package works through repository-based marketplace configuration, but it is not yet a one-click public-directory install.
Prompts to Try
- Find five business ideas for independent accountants who spend too much time on client follow-up. Check the evidence for the strongest three.
- Validate a service that monitors AI-agent tool calls for data leakage. Separate direct evidence from adjacent evidence.
- Compare the best two developer-tool opportunities for a two-person bootstrapped team. Show the main assumption behind each.
- Find business ideas using AI, but exclude opportunities without inspectable supporting evidence.
These prompts work because they give the agent a customer, problem, or operating constraint. A broad request such as “give me startup ideas” can still return results, but the comparison becomes more useful when the agent knows what you can build and who you want to serve.
What the Plugin Can and Cannot Establish
Trend Seeker can help an agent find relevant opportunities, retrieve stored source excerpts, compare research summaries, and expose gaps. It cannot establish market size, willingness to pay, founder fit, or whether your version of a product will succeed.
Use the plugin to build a research shortlist. Then open the original sources, interview the target users, and test for commitment. The market-signal research guide covers that wider process, and the Opportunity Score methodology explains how Trend Seeker prioritizes evidence without turning the score into a forecast.
Frequently Asked Questions
Can Claude Code find business ideas with Trend Seeker?
Yes. The plugin gives Claude Code workflows and read-only MCP tools for finding opportunities, validating an existing idea, comparing a shortlist, and retrieving supporting evidence.
Can Codex use Trend Seeker for business idea research?
Yes. The public package includes a Codex plugin manifest, shared research skills, and the hosted Trend Seeker MCP connection. The repository can be referenced by a configured Codex marketplace.
What is a market research MCP server?
A market research MCP server exposes structured research tools to an AI client. Trend Seeker's server lets supported agents search business ideas, retrieve idea details and evidence, browse categories, and compare opportunities.
Does the Trend Seeker plugin prove that a business idea will work?
No. It retrieves research signals and stored public evidence. Scores help prioritize further research, but they do not forecast revenue, market size, or startup success.
Ready to find your next business idea?
Explore validated business ideas backed by real user demand.