Capability intelligence for job ads,
for job boards, ATS and hiring agents.
Send a job ad. Get back a structured read of what the role really asks for: true seniority, the capability mix, how inflated the requirements are, and the engagement type. It reads roles, never people.
Read between the lines to see the bigger picture. The name is ReLoop.me, always with .me. Every read follows the Fair Read Charter.
Role Quick Check · free, no key, about 15 seconds
Thinking of applying, or building an agent that applies? Check the ad first: the level the scope really supports, whether the title matches it, how inflated the requirements are, whether pay is disclosed, the core capabilities, and the shape of the role on 5 axes. Ads in English, Romanian, French, German, Spanish, Italian, Dutch, Polish, Czech, Hungarian and Portuguese are read in their own language. It never estimates pay.
The ad text is sent to ReLoop.me's reader. Emails and phone numbers are removed before the read. Limits: 5 quick checks per visitor and 50 in total per day.
A read of the role,
not of the people who apply.
The ReLoop.me Role Intel API is a capability intelligence API for job ads. Capability intelligence means reading what a role or a person actually demands or shows (scope, level, capabilities and how the work is set up) from the evidence, rather than from titles or keywords. ReLoop.me reads roles and people on the same axes, for skills-based hiring and nonlinear careers.
Built for job boards, ATS and hiring agents that want skills-based reads of roles. Outputs may not be used to screen, rank or reject candidates.
Quick check, API full read,
full Role Intel report.
Cheapest first. Level, title vs scope and requirement inflation come from the same shared read in all three; the paid reads add the evidence behind each answer.
| Role Quick Check | API full read | Role Intel report | |
|---|---|---|---|
| Price | Free, no key | €3 per delivered read | €7 per ad, one-time |
| Time | About 15 seconds | Usually 5–8 minutes; a pilot read held for review, up to one working day | Within 24 hours, by email |
| Input | Job ad text, 300–15,000 characters | Job ad text, 200–20,000 characters | A job ad, through the order form |
| Languages | 11, each ad read in its own language | ||
| Level | The level the scope supports, marked indicative | The same level, with the gap to the title and the evidence | The same, explained in prose |
| Title vs scope | Aligned, higher, lower or unclear | The same verdict, with the reasons | The same, explained |
| Capabilities | Up to 3 shown (the full list is in the JSON) | Capability mix, weighted, with evidence from the ad | Full capability map with evidence |
| Requirement inflation | Level and codes | Index with the claims behind it | Claims vs evidence, line by line |
| Shape of the role | Up to 5 axes | Up to 10 axes, including signals about the ad itself | The same axes, explained |
| Pay | Whether the ad shows pay (yes, partial, no); no estimate | No pay | Estimated range, for covered markets |
| AI exposure, operating stage, market | No | Yes | Yes |
| Career backgrounds that transfer in (types, never actual people) | No | Named only | Yes |
| Sourcing, how the ad lands, transferability, calibration questions, CV lines | No | No | Yes |
| Output | JSON (schema role-quick-check/1.3) and an embeddable block | JSON (Role-Demand Brief) | Web report and email |
| Daily limits | 5 per visitor, 50 in total; keys get their own allowance | A daily cap per key, set with you | None |
| Stored | The ad text is not kept; length, a fingerprint and the result fields are logged | Cleaned ad and read, 3 months | 3 months |
| Human review | No | Pilot reads can be held for a person's check | Automated quality check on a human-calibrated framework |
| Fair Read rules | Travel with every read, as a reading_rules block | ||
Three calls,
one structured read.
1Your key
Every request carries your key in the header x-reloop-key. Get a free sandbox key below: it returns a sample read instantly and costs nothing. Your live key (rl_live_…) arrives by email with your first top-up.
2Submit a job ad
curl -X POST https://api.reloop.me/role-intel/v1/reads \
-H "x-reloop-key: YOUR_KEY" \
-H "content-type: application/json" \
-d '{"jd_text": "…full job ad text…", "location": "Bucharest", "client_ref": "your-job-id-123"}'Response 202: {"read_id":"rr_…","status":"processing","poll_url":"…"}
| Field | Required | Notes |
|---|---|---|
jd_text | yes | The full ad, 200–20,000 characters. Email addresses and phone numbers are removed before storage. |
location | no | City or country. The engine detects it if missing. |
client_ref | no | Your own ID, returned with the read. |
callback_url | no | An https URL. We POST {read_id,status} when the read is done. |
3Get the read
curl "https://api.reloop.me/role-intel/v1/reads?read_id=rr_…" -H "x-reloop-key: YOUR_KEY"Status moves from processing to delivered, usually in 5–8 minutes. Poll every 60 seconds or use a callback. During a pilot, the first reads can show held (review queue) until a person at ReLoop.me has checked them, usually within one working day.
Held reasons (held_reason): review_queue (pilot reads, released after a human check); programme_overview (the text is a programme page that points to separate role ads: send the ad for one pathway or position); quality_review_hold or delivery_gate_block (a quality check stopped the read). Held reads are not charged, except review-queue reads once released.
Fields in the read
true_seniority | The level the scope supports (Emerging → Executive), the gap with the stated title, and why. |
|---|---|
capability_mix | The capabilities the role demands, weighted primary / secondary / supporting, each with evidence from the ad. |
jd_inflation_index | none / low / medium / high, with the signals behind it. |
scope, engagement_type | Decision rights, accountable surface, and employment / contract / fractional / board. |
operating_style | How the role is set up to work, on up to 10 axes from −1 to +1: execution ↔ strategy, maintain ↔ build, specialist ↔ generalist, direct authority ↔ matrix influence, clear ↔ ambiguous scope, local ↔ international, senior IC ↔ people manager, technical depth ↔ translation. Axes marked ad_signal (candidate-friendly ↔ employer-risky, pay transparent ↔ opaque) read the ad, not the work. Neither end is better. |
ai_exposure | How much the role works with AI (e.g. User, Builder), with one line on what that means here. |
operating_stage | The stage of the function or company the role sits in (founding, scaling, steady state). |
market | The resolved location band (e.g. RO_iasi) and whether it was stated or inferred. |
links, full_read_includes | Links to ReLoop.me pages you can show next to the read, and the names of the sections only the full Role Intel read carries. |
reading_rules | The rules every read travels with (see Reading rules): the charter version, the seven rules, what the data may not be used for, and how to ask for a re-read. |
engine | The engine versions that produced the read. Every read names its versions. |
Example response (excerpt of a real read, employer name removed)
{
"read_id": "rr_...",
"status": "delivered",
"rdb": {
"rdb_version": "0.1.3",
"role_title_stated": "AI Lead",
"true_seniority": {
"level": "Director",
"track": "people_manager",
"title_stated_gap": "The ad describes Director or Head-of-AI scope: CoE establishment, group-wide mandate, executive reporting, budget and vendor management. The title reads as senior IC or team lead."
},
"capability_mix": [
{ "capability": "AI production delivery: moving PoCs into embedded engineering practice",
"weight": "primary", "level": "Expert",
"evidence": "Track record delivering AI capabilities into production" }
],
"jd_inflation_index": { "rating": "high", "signals": ["'owns this journey end to end' with no team, budget or decision rights named", "..."] },
"engagement_type": "employment",
"operating_style": { "scale": "-1 = fully the low end, 0 = balanced, +1 = fully the high end",
"axes": [ { "axis_id": "execution_strategy", "low_end": "Execution", "high_end": "Strategy", "position": -0.2,
"label": "Strategy-setting is named first, but delivery and ROI reporting weigh equally" }, "..." ] },
"pay_status": "not_in_api",
"reading_rules": { "charter": "https://www.reloop.me/charter", "not_allowed": ["ranking or filtering candidates", "..."] }
}
}Reads are automated and AI-generated content, labelled as such (EU AI Act Art. 50). The API reads job ads and does not evaluate candidates. Values are estimates, not legal advice; null means not measured, never zero. The employer is referred to as "the employer". Outputs may not be used to screen, rank or reject candidates. Please credit "Role-Demand Brief by ReLoop.me" when you show a read to your users.
Free sandbox key,
by email, in about a minute.
No card needed. The sandbox key returns a sample read instantly. Sandbox keys allow 5 quick checks a day.
Pay per delivered read
| Sandbox key | Free · sample read · no card |
|---|---|
| Core read | €3 per delivered read |
| Pay | Not part of the API. Estimated pay ranges are in the full Role Intel report. |
| Prepaid balance | Top up by card: €25, €100 or €250 (links in your sandbox email). Reads held by a quality gate or that fail are not charged. |
Final price, no VAT added (ZORRI S.R.L. is not VAT-registered). Invoice by email. Volume, custom calibration or invoicing in advance: hello@reloop.me.
MCP server,
three tools.
Machine-readable spec: OpenAPI 3.1. Agent summary: llms.txt.
quick_check_job_ad: the quick check above, in about 15 seconds. No key needed.read_job_ad: submits a full read (needs a key) and returns aread_idat once.get_read: fetches the full read byread_idwhen it is ready (5–8 minutes).
The Fair Read rules travel in the tool descriptions.
Add it to your MCP config:
{
"mcpServers": {
"reloop-role-intel": {
"command": "npx",
"args": ["-y", "reloop-role-intel-mcp"],
"env": { "RELOOP_API_KEY": "rl_sandbox_... (optional, only for full reads)" }
}
}
}The rules travel
with the data.
Every read carries a reading_rules block from the ReLoop.me Fair Read Charter (v0.1). If you build on ReLoop.me data, keep the block with the data and keep the rules.
- Evidence over titles. Read what the work demands or shows, not what it is called.
- A gap is not a deficit. Career breaks, caregiving, moves and changes of field are never scored down.
- Unknown is not zero. A missing value stays unknown. Never fill it with an average, a guess or a penalty.
- One ruler for both sides. Roles and people are read on the same axes, by a reader neither side pays to lean its way.
- The subject owns the read. For a role, that is the client who sent the ad. For a person, the person.
- Describe, don't rank. A read places a role or a person on axes. It never orders people, filters them out or decides who gets through.
- Show the why. Every judgement carries its reasoning in plain words.
Not allowed: ranking or ordering people, screening or rejecting people, inferring personality or protected traits, imputing missing values. To contest a read, write to hello@reloop.me. AI agents: see llms.txt.
The API does not give salary ranges: not in the quick check, not in the full read. Estimated pay ranges are part of the full Role Intel report, for a limited set of markets where ReLoop.me has calibrated pay data. They are estimates, never offers. Platforms that need pay data can ask about a pay add-on on a paid plan: hello@reloop.me.
The full Role Intel read of the same ad adds: the career backgrounds and capability patterns that transfer into the role (types, never actual people) and where the role transfers; how the ad lands with candidates and with screeners; transferability by industry, function, seniority and geography; calibration questions; pay scenarios and comp structure; and positioning lines for a CV. See the Role Intel sample.
Job ad text only.
No CVs, no candidate data.
- What you send: job ad text only. Email addresses and phone numbers are removed before anything is stored.
- Quick check: the ad text is not stored; only its length, a fingerprint and the result fields are logged, to count use.
- Full reads: the cleaned ad and the read are kept for 3 months to answer questions, re-reads and billing, then deleted. Ask for earlier deletion at hello@reloop.me.
- Limits: quick check 5 per visitor and 50 in total per day; sandbox keys 5 quick checks per day, live keys 20; full reads have a daily cap per key (write to raise it).
- Pilot service: best-effort availability, versioned outputs (each read names its engine and schema versions), changes announced on this page.
- Re-reads: if a read looks wrong, write to hello@reloop.me with the read_id. A person at ReLoop.me looks at it within 2 working days.
401 | Key unknown or inactive |
|---|---|
402 | Balance too low |
413 | Ad too long |
422 | Ad missing, too short, or not a job ad |
429 | Daily limit reached |
502 | The read could not be completed; try again after retry_after seconds |
| Role Intel | The full read of one job ad, for a candidate or a hiring team. |
|---|---|
| Profile Intel | The same capability lens applied to a CV: what a profile really shows, where it transfers, and its market value. Useful to point candidates to from a job page. |
| For hiring teams | Pilots for EU pay transparency and skills-based hiring. |
| Methodology | How reads are produced, and what the engine will not do. |
| Free guides | How to read a job ad · EU pay transparency · Glossary |