Search my portfolio

Search cases, decisions, results and experience.

Assistant access

Bring the evidence
into your assistant.

Setup guidance, dated client observations and the source checks behind a useful conversation.

Use my published work in your AI assistant

Map an opportunity to published evidence, prepare a substantive conversation, or inspect project maturity. You can also browse the sources without an AI account.

Connection details

Server URL: https://olasobo.com/api/public-evidence-mcp

Choose no server authentication. Expect two read-only tools: search and fetch. Your assistant may require an account, subscription, consent or organization approval.

Search discovers approved sources using topic keywords. Fetch the full record before making substantive claims, and retain contribution scope, maturity and limitations. If search is empty, make at most one meaningful refinement. Empty results do not establish that evidence is absent elsewhere.

ChatGPT

On ChatGPT web, eligible Pro accounts can connect read/fetch MCPs in developer mode. On Business, an administrator or owner creates the custom app; members can use it after workspace publication. On Enterprise and Edu, members with the required workspace permissions may create it. Follow OpenAI's current setup instructions for your plan.

Enable developer mode through the settings available to your account. In Apps, choose Create, enter the server URL, select no authentication, scan the tools and create the app. Select it in a new chat. Select it again when a follow-up needs new retrieval. Organization publication requires an authorized administrator or owner; a personal connection is separate from a public directory listing.

Try: “Use Ola's published evidence to prepare three questions about production product leadership. Fetch the relevant records and cite their limits.”

Changes to tool definitions may require a refresh or recreation through your workspace's controls.

Claude

Claude documents custom remote connectors on Free, Pro, Max, Team and Enterprise, with one custom connector on Free. The documented individual setup steps cover Pro and Max; Free setup has not been tested here. Check Claude's current instructions.

For Pro or Max, open Customize → Connectors → Add → Add custom connector. Enter a name and the server URL, continue, review the detected authentication and choose No sign in. Leave request headers empty. Add the connector, then enable it for the conversation through the chat's Connectors menu.

For Team, an Owner or Primary Owner adds the connector through Organization settings → Connectors. Enterprise also permits custom roles with “Manage access to Libraries.” Members can then connect through Customize → Connectors. Remote requests originate from Anthropic's cloud.

Try: “What is MIN–MAX Living's published maturity? Fetch the evidence and distinguish capabilities from unestablished adoption or commercial outcomes.”

Gemini

Google currently documents custom apps for adults in the US using a personal Google Account, English, and Keep Activity enabled. Setup begins on Gemini web; work and school accounts are excluded. Review Google's current instructions.

Open Settings → Connected Apps, or Personal Intelligence → Connected Apps if shown. Under Custom apps, add the server URL, choose Next and review consent. This server requires no credentials. Confirm that search and fetch are listed. In a fresh prompt, type @ and select the app.

Try: “Use Ola's published enterprise evidence. Fetch the relevant records, distinguish product leadership from engineering ownership, and preserve their limitations.”

Cursor for technical evaluators

Use a Cursor account and installation permitted by your organization. Create .cursor/mcp.json in your project with this remote entry:

{ "mcpServers": { "ola-public-evidence": { "url": "https://olasobo.com/api/public-evidence-mcp" } } }

Enable the server and its tools in Customize, then use Agent chat. Follow Cursor's current MCP documentation and your team's policies.

Try: “Find the published workflow platform migration evidence, fetch the record and distinguish migration scope from production operation.”

What has been verified

Documentation checked: 4 October 2026. The endpoint and Gemini observations below were recorded on 2 October 2026.

An anonymous HTTP test retrieved sources from the public endpoint. That test establishes endpoint retrieval separately from each assistant's connection and onboarding.

Recorded client observations and limits
ClientRecorded observationRemaining limitation
ChatGPT webHistorical owner connectionFresh non-owner onboarding and a repeated task were not verified
Claude webSetup documentation checkedNo account was available for execution
Gemini Spark webNative connection and one task showed search and two fetch actionsNon-owner account classification was unresolved; a second task did not complete
CursorSetup documentation checkedNo client execution was performed

The Gemini answer preserved contribution and evidence limitations. Its citation opened Google search; opening the canonical source directly worked. This was a visible client action trace, without a captured wire transcript. Client build numbers were unavailable. These observations do not establish reliable completion across accounts or assistants.

Sources and troubleshooting

Open the canonical URL returned by fetch and inspect the linked passage. For contact, try “How can a recruiter contact Ola about a Lead Product Manager opportunity?” If search is empty, refine once to “contact” and open Contact Ola. In the recorded public-endpoint test on 2 October, the full question returned no result; the shorter refinement found the contact source.

If tools are missing, check account eligibility, organization policy, the server URL and whether the app or connector is enabled or selected in your assistant. Refresh stale definitions through your assistant's documented controls. If connection fails, record the client, date and error; do not add a token to the URL. Ordinary source pages remain available.

These sources describe my work, contribution and product outcomes. Fetch the relevant record and follow its linked case for context. Keep private information out of public answers.

External assistants control their own synthesis. Server withdrawal stops future retrieval but cannot erase copies already in a client's chat.

Continue the story

Access to a source is the beginning of the answer.

I expose evidence for assistants to retrieve. My evaluation work examines the next step: whether the answer preserves the requested case, contribution and measurement context.

Inspect answer quality and the release decision Explore the cases directly Discuss a role