# Bootstrap prompt

Paste the prompt below into an agent that can reach the Metix AI Platform: one with the [metix-skills](https://github.com/MetixAI-Official/metix-skills) installed, the MCP server connected, or plain REST access with `METIX_KEY` set. Followed end to end for a company with about 110 open postings, it costs 28 to 35 API Credits, most of it to read the postings. It stops and asks before 50.

```text
Tell me what Notion is hiring for, using the Metix AI Platform. Work only through the public Platform (REST at https://mira-api.metix.ai, the MCP server, or the metix-skills) with the key in METIX_KEY, and never print the key.

1. Read before querying. Call GET /contract (free) and read the job fields company.name, is_open, title, location.city and seniority. Check the balance with GET /auth/key/status (free) at the start and at the end, and stop and ask before the run passes 50 API Credits.

2. Population. Job postings with company.name eq "Notion" and is_open eq true, in one all node.

3. Count first. Send the filter with size 1 and tell me the total and what reading them all will cost (1 API Credit per 5 postings read, plus about 1 per 25 ids searched). If that is over the ceiling, ask me before reading.

4. Budget. Collect the ids with size 100 pages, then read the postings in batches of 100 with only title, location.city, location.country and seniority.

5. Audit. Check that every posting really belongs to this company and not to a namesake. List any that look wrong and leave them out.

6. Clean. Keep reposts as separate postings and say how many titles appear more than once.

7. Group. Put each title into one family by keywords, first match wins, in this order: data and AI, engineering, product and design, sales and partnerships, marketing, customer and education, finance, legal and people. Show me the keyword lists and the titles that fell into no family, and fix the lists until none are left over.

8. Outputs. Three tables with counts: by family, by city, and by seniority, with the snapshot date. Keep the records to our conversation.

9. Chart. One horizontal bar chart of the families, largest first, with the count on each bar. Title it with what stands out.

10. Limits. Say that this is one day of open postings, that some roles are never posted and some postings outlive the hire, and that keyword families are coarse.
```

## Adapt it

| To change | Edit | For example |
| --- | --- | --- |
| The company | Step 2 | "Figma", "Databricks" (check the total first) |
| The families | Step 7 | Split engineering into platform, product and security |
| Recent roles only | Step 2 | Add posted_date gte now-30d |

## What to ask before running it

1. Which company, spelled as it appears on its postings?
2. All open roles, or only those posted recently?
3. Which families matter for your question?