unclecode/crawl4ai · error · HTTPException

Token issuance is disabled: no api_token is configured on th

Error message

Token issuance is disabled: no api_token is configured on the server.

What it means

An explicit 403 raised by POST /token when the server has no security.api_token configured. Token issuance fails closed: previously the endpoint minted a JWT to anyone whose email domain merely had an MX record, so in the current design the absence of a configured api_token disables issuance entirely rather than falling back to weak verification.

Source

Thrown at deploy/docker/server.py:540

@app.exception_handler(Exception)
async def _unhandled_exception_handler(request: Request, exc: Exception):
    cid = _uuid.uuid4().hex[:12]
    logger.exception("unhandled exception [cid=%s]", cid)
    return JSONResponse(
        {"error": "Internal server error", "correlation_id": cid},
        status_code=500,
    )


# ──────────────────────── Endpoints ──────────────────────────
@app.post("/token")
async def get_token(req: TokenRequest):
    expected_token = config.get("security", {}).get("api_token", "")
    if not expected_token:
        # Fail closed: without a configured api_token the old behavior minted a
        # JWT to anyone whose email merely had an MX record. Refuse instead.
        raise HTTPException(
            403,
            "Token issuance is disabled: no api_token is configured on the server.",
        )
    if not req.api_token or not constant_time_eq(req.api_token, expected_token):
        raise HTTPException(401, "Invalid or missing api_token")
    if not verify_email_domain(req.email):
        raise HTTPException(400, "Invalid email domain")
    token = create_access_token({"sub": req.email})
    return {"email": req.email, "access_token": token, "token_type": "bearer"}


@app.post("/config/dump")
async def config_dump(
    data: dict,
    _td: Dict = Depends(token_dep),
):
    try:
        return JSONResponse(_config_from_json(data))

View on GitHub (pinned to 7e80152142)

Solutions

  1. Configure security.api_token in the server config (via its config file or the secret-injection mechanism the deployment uses), then restart/reload the server.
  2. Verify with a config-health check or by confirming the secret is mounted before calling /token.
  3. If email-domain-based open issuance was intended, that behavior is intentionally removed - use the configured api_token flow.

Example fix

# before
# server config (yaml/json) lacks:
#   security: {}
post(f"{base}/token", json={"email": "a@b.com", "api_token": ""})  # 403

# after
# server config:
#   security: {"api_token": "<shared-secret>"}
post(f"{base}/token", json={"email": "a@b.com", "api_token": "<shared-secret>"})
Defensive patterns

Strategy: validation

Validate before calling

def token_issuance_available(base) -> bool:
    """Probe whether the server has an api_token configured.
    403 from /token means issuance is disabled."""
    r = post(f"{base}/token", json={'email': 'probe@example.com', 'api_token': 'x'})
    return r.status_code != 403

Try / catch

try:
    tok = post(f"{base}/token", json={'email': email, 'api_token': secret})
except HTTPError as e:
    if e.response.status_code == 403:
        raise RuntimeError(
            'server has no security.api_token configured; ask the operator to set it'
        ) from e
    raise

Prevention

When it happens

Trigger: Calling POST /token with any credentials on a server whose config lacks the security.api_token key (fresh install, default config, or a deployment that intended JWT-only auth without shared-secret bootstrapping).

Common situations: New deployments that never set the api_token secret; environments where the token was expected to be injected via env/secret store but was not (secret mount missing); staging servers sharing a config template that omits the security block.

Related errors


AI-assisted analysis of unclecode/crawl4ai@7e80152142 (2026-08-14). Data as JSON: /api/errors/87a978f50a8ef2f0. Report an issue: GitHub.