Agentic Evidence

API and workflow

Paid teams create assessments, generate candidate sessions, then review reports.

POST /api/assessments
Authorization: Bearer ae_key_...
{
  "title": "Fix checkout rounding bug",
  "repoUrl": "https://github.com/company/candidate-repo",
  "instructions": "Find and fix the tax rounding bug. Add tests."
}
POST /api/assessments/:id/session
Authorization: Bearer ae_key_...
{
  "candidateEmail": "candidate@example.com",
  "candidateName": "Alex Chen",
  "role": "Senior Backend Developer",
  "dueAt": "2026-05-01T17:00:00Z"
}
POST /api/workspaces
Authorization: Bearer ae_key_...
{
  "assessmentId": "asmt_...",
  "candidateEmail": "candidate@example.com",
  "candidateName": "Alex Chen",
  "role": "Senior Backend Developer"
}
POST /api/vm/provision
{
  "sessionId": "sess_...",
  "token": "cand_..."
}
POST /api/vm/teardown
{
  "sessionId": "sess_...",
  "token": "cand_..."
}
curl -fsSL "https://agentic-evidence.fly.dev/vm/bootstrap.sh?sessionId=sess_...&token=cand_..." | bash
ae-codex
ae-claude
ae-snapshot
POST /api/vm/events
{
  "sessionId": "sess_...",
  "recorderToken": "rec_...",
  "type": "agent_transcript",
  "payload": { "text": "terminal transcript..." }
}
GET /github-action.yml

# Save this into the candidate repo as .github/workflows/agentic-evidence.yml.
# Add repository secrets:
# AGENTIC_EVIDENCE_SESSION_ID=sess_...
# AGENTIC_EVIDENCE_TOKEN=cand_...
POST /api/github/evidence
{
  "sessionId": "sess_...",
  "token": "cand_...",
  "repository": "company/candidate-repo",
  "sha": "abc123",
  "testCommand": "npm test",
  "testStatus": "0",
  "testOutput": "4/4 passed",
  "diff": "git diff output"
}
GET /report/:sessionId?token=rpt_...
or
GET /report/:sessionId
Authorization: Bearer ae_key_...

Guardrails

6000

Max prompt characters.

700

Max output tokens per model response.

500

Monthly agent messages.

2,000,000

Monthly estimated tokens.

72h

Candidate session TTL.

240m

VM teardown target.