Eaglecross API & MCP

Authorization: Bearer ec_live_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

claude mcp add --transport http eaglecross https://api.eaglecross.xyz/mcp

gemini mcp add --transport http eaglecross https://api.eaglecross.xyz/mcp

{
  "mcpServers": {
    "eaglecross": {
      "url": "https://api.eaglecross.xyz/mcp",
      "headers": { "Authorization": "Bearer ec_live_…" }
    }
  }
}

Research these 5 founders and tell me who to email this week, with the opener for each:
- Jane Doe, CEO, Acme (acme.com)
- ...

---
name: eaglecross
description: Research B2B leads with the Eaglecross API - find recent buying signals (hiring, funding, launches, news, tech-stack changes), score each lead against an ICP preset, and get a personalized cold-email subject and opening line for leads worth contacting. Use when the user wants to research prospects, find who to email, or write icebreakers. Works in Claude Code, Cursor, Codex, Gemini CLI and any agent that can make HTTP requests.
---

# Eaglecross API

Eaglecross turns a list of prospects into a ranked outreach list. For each lead it researches the company's
website, news, job posts and tech stack, scores the lead against the user's Ideal Customer Profile (ICP) preset,
flags it `reach` (contact now) or `skip`, and writes an email subject + opening line for reach leads.
It is the same engine, credits, presets and templates as the web app (https://app.eaglecross.xyz).

Docs: https://www.eaglecross.xyz/api · Interactive reference: https://api.eaglecross.xyz/docs · OpenAPI: https://api.eaglecross.xyz/openapi.json

## Auth

- Base URL: `https://api.eaglecross.xyz`
- Header on every request: `Authorization: Bearer $EAGLECROSS_API_KEY`
- Keys start with `ec_live_`. The user creates one in the web app under Settings → API & MCP (every plan).
  Never print or commit the key; read it from the environment.
- Prefer MCP when the client supports it: `https://api.eaglecross.xyz/mcp` (OAuth sign-in, no key needed).

## Workflow

1. `GET /v1/account` - check `credits_balance` before starting. Tell the user the cost first.
2. `POST /v1/jobs` with `{"name", "preset"?, "lookback_window"?, "advanced_analysis"?, "leads": [...]}`
   - each lead: `company` and `website` required; `name`, `title`, `linkedin_url`, `company_description`,
     `location`, `industry` optional (name + title make better icebreakers)
   - or `POST /v1/jobs/csv` (multipart `file`, max 5 MB) with the web app's CSV columns
3. Poll `GET /v1/jobs/{id}` every 15-30 s until `status` is `done` (a batch takes ~30-90 s).
   `stalled: true` means no progress for 10+ minutes: tell the user to contact support@eaglecross.xyz.
4. `GET /v1/jobs/{id}/leads?reach_flag=reach` for leads to contact (highest score first, `limit` ≤ 200).
   `GET /v1/jobs/{id}/leads.csv` for everything as CSV, or `POST /v1/jobs/{id}/export-link` for a 1-hour link.
5. `GET /v1/leads/{id}` for every signal with its source URL and date.
6. `POST /v1/leads/{id}/icebreaker` to rewrite an opener (`instructions`, `template`, `subject_template`).

Other endpoints: `GET /v1/presets`, `GET /v1/jobs` (status, limit ≤ 100, offset), `POST /v1/jobs/{id}/cancel`,
`DELETE /v1/jobs/{id}`, `GET /v1/settings`, `PATCH /v1/settings`.

Options: `preset` = `lead_gen_agency` | `b2b_saas` | `ai_agency` | `custom` (default: the user's saved preset);
`lookback_window` = `aggressive` (fresh signals only) | `balanced` | `lenient`; `advanced_analysis` (Pro only).

## Reading results

- `reach_flag`: `reach` = strong recent signal, contact now; `skip` = nothing strong enough yet (no icebreaker).
- `score`: 0-100 after time decay. `reach_reason`: why. `intent_summary`: what the company is focused on.
- `icebreaker_subject` / `icebreaker_body`: ready to paste into an email.
- `top_signals` / `signals`: `type`, `snippet`, `source_url`, `published_date`, `tier`, `score`.
  Cite the source URL when you tell the user why a lead is worth contacting.

## Credits and limits (same as the web app)

- 1 credit per lead, including its icebreaker; 2 with advanced analysis (Pro). 1 credit per standalone rewrite.
  Reads are free. Failed leads are refunded.
- Starting a job needs at least 1 credit per lead. There is no other cap on leads per job.
- One job processing at a time per account (web app and API combined). Leads inside a job run in parallel
  per plan: Free 2, Standard 5, Pro 10. Signals kept per lead: Free 8, Standard 16, Pro 20.
- Over MCP, `research_leads` takes up to 100 leads, or a pasted CSV of up to 500 rows, per call.

## Safety

- Confirm with the user before researching more than 20 leads (it spends credits) and before
  `DELETE /v1/jobs/{id}` or `POST /v1/jobs/{id}/cancel`.
- Don't start a second job while one is processing; you'll get 429 `too_many_active_jobs`.

## Errors

Shape: `{"error": {"code": "...", "message": "...", "details"?: [...]}}`

| Status | code | Do |
|---|---|---|
| 401 | unauthorized | Key missing, invalid, expired or revoked |
| 402 | insufficient_credits | Ask the user to add credits at https://app.eaglecross.xyz/pricing |
| 403 | plan_required | Feature needs a higher plan (advanced analysis = Pro) |
| 404 | not_found | No such job or lead in this account |
| 409 | account_not_ready | User must sign in to the web app once |
| 413 / 400 | file_too_large / invalid_csv | Fix the CSV as the message says |
| 422 | invalid_request / no_signals | See `error.details` |
| 429 | too_many_active_jobs | Wait for the running job to finish |
| 502 / 503 | job_start_failed / upstream_unavailable | Retry with backoff; nothing was charged |

{ "error": { "code": "insufficient_credits", "message": "This job needs at least 40 credits ..." } }