{"id":"CVE-2026-55538","aliases":["GHSA-r7v3-x45f-g7hp","PYSEC-2026-3894"],"title":"PraisonAI: [Auth Bypass] `praisonai serve agents --api-key` is silently ignored — agent-invocation routes (`POST /agents`, `POST /agents/…","summary":"PraisonAI: [Auth Bypass] `praisonai serve agents --api-key` is silently ignored — agent-invocation routes (`POST /agents`, `POST /agents/{agent_name}`) run unauthenticated","severity":"high","cvss":7.3,"cvssVector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:L","vendor":"praisonai","product":"praisonai","ecosystem":"pip","affected":["praisonai < 4.6.58"],"patched":["praisonai 4.6.58"],"published":"2026-08-25","updated":"2026-09-10","sourceUpdated":"2026-09-10T12:25:41.490935690Z","source":"OSV","sourceUrl":"https://osv.dev/vulnerability/GHSA-r7v3-x45f-g7hp","references":[{"url":"https://github.com/MervinPraison/PraisonAI/security/advisories/GHSA-r7v3-x45f-g7hp"},{"url":"https://github.com/MervinPraison/PraisonAI/commit/2f9677abb2ea68eab864ee8b6a828fd0141612e1"},{"url":"https://github.com/MervinPraison/PraisonAI"},{"url":"https://github.com/MervinPraison/PraisonAI/releases/tag/v4.6.58"},{"url":"https://pypi.org/project/praisonai"},{"url":"https://github.com/advisories/GHSA-r7v3-x45f-g7hp"},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-55538"}],"tags":["osv","pip","nvd","ghsa"],"epss":0.00258,"epssPercentile":0.17758,"cwe":["CWE-306"],"ingestedAt":"2026-08-25T15:27:53.381Z","slug":"CVE-2026-55538","body":"## Overview\n\n### Summary\n`praisonai serve agents` exposes HTTP routes that invoke registered agents. The CLI advertises `--api-key` with help text \"API key for authentication\", parses it, and forwards it into `ServeHandler`. But `_create_agents_app()` **never reads `config[\"api_key\"]` again** and installs no auth dependency or middleware on its direct routes. The configured key is a no-op flag.\n \nAs a result, an unauthenticated network caller can invoke exposed agents (`POST /agents` and `POST /agents/{agent_name}`) even when the operator passed `--api-key`. Requests with no credentials, a wrong `Authorization: Bearer`, a wrong `X-API-Key`, or an empty bearer all reach `agent.start()`.\n \nThe failure is made sharper by the fact that a **working auth dependency already exists in the same module** — `praisonai.api.agent_invoke.verify_token` guards every `/api/v1/...` route with `Depends(verify_token)` and is mounted into the very same app. The direct n8n-compat routes simply do not use it.\n\n## Technical Detail\n \n### Source-to-sink trace\n \n**1. CLI advertises and forwards `--api-key`:**\n \n```python\n# cli/commands/serve.py\n@app.command(\"agents\")\ndef serve_agents(..., api_key: Optional[str] = typer.Option(None, \"--api-key\", help=\"API key for authentication\")):\n    ...\n    if api_key:\n        args.extend([\"--api-key\", api_key])\n    exit_code = handle_serve_command(args)\n```\n \n**2. `cmd_agents()` parses `api_key` into the spec — and that is the last time it is touched:**\n \n```python\n# cli/features/serve.py — cmd_agents()\nspec = { ..., \"api_key\": {\"default\": None} }\nparsed = self._parse_args(args, spec)\napp = self._create_agents_app(parsed)\n```\n \nA grep of the entire `cli/features/serve.py` for `api_key` returns **only** the two `spec` entries (`cmd_agents` line ~199 and `cmd_unified` line ~847). `config[\"api_key\"]` is never read inside `_create_agents_app()` / `_create_unified_app()`; it is never compared, and no dependency is attached.\n \n**3. `_create_agents_app()` imports `FastAPI, HTTPException, Request` — no `Depends`, no `Header`, no auth middleware.** Every `HTTPException` raised in the agents routes is `400`/`404`/`500` (validation / not-found / execution error); none is `401`.\n \n**4. Sink — unauthenticated request reaches `agent.start()`:**\n \n```python\n# cli/features/serve.py\n@app.post(\"/agents/{agent_name}\")          # n8n compatibility route\nasync def invoke_single_agent(agent_name: str, request: Request):\n    body = await request.json()\n    query = body.get(\"query\", \"\") or body.get(\"message\", \"\")\n    ...\n    agent = agent_invoke.get_agent(agent_name)\n    result = await loop.run_in_executor(None, agent.start, query)   # no auth anywhere above\n    return {\"response\": str(result)}\n \n@app.post(path)                            # default path \"/agents\"\nasync def invoke_agents(request: Request, query_data: AgentQuery = None):\n    ... agent.start(query) ...\n```\n\n### The auth dependency exists — it just isn't applied here\n \n`_create_agents_app()` mounts the `agent_invoke` router into the same app:\n \n```python\n# cli/features/serve.py\nif getattr(agent_invoke, 'FASTAPI_AVAILABLE', False) and hasattr(agent_invoke, 'router'):\n    app.include_router(agent_invoke.router)\n```\n \nThat router properly authenticates every sensitive route:\n \n```python\n# api/agent_invoke.py\nCALL_SERVER_TOKEN = os.getenv('CALL_SERVER_TOKEN')\nasync def verify_token(request, authorization=Header(None)) -> None:\n    ...\n    if token != CALL_SERVER_TOKEN:\n        raise HTTPException(status_code=401, detail=\"Unauthorized\")\n \n@router.get(\"/api/v1/agents\")\nasync def list_agents(_: None = Depends(verify_token)): ...   # and register/unregister/info all use it\n```\n \nSo in the same process `GET /api/v1/agents` returns `401` without a token, while `POST /agents/{agent_name}` returns `200`. Note also that `verify_token` reads the `CALL_SERVER_TOKEN` env var — **not** the CLI `--api-key` — so the CLI option feeds no auth path at all.\n \n### Trigger conditions\n \n```\npraisonai serve agents --file agents.yaml --host 0.0.0.0 --port 8765 --api-key expected-secret\nPOST /agents/{agent_name}   body {\"query\":\"...\"}   with no / wrong / empty credentials\n```\n \n## Proof of Concept\n \nBuilt the **real** `_create_agents_app()` and exercised it over HTTP via FastAPI `TestClient`. Only `praisonaiagents.Agent` is stubbed (`.start()` returns `EXEC:<query>`), so no real LLM/credentials. `CALL_SERVER_TOKEN=expected-secret` was set so the sibling `/api/v1` router is genuinely armed — making the contrast explicit.\n \n```\nOperator started with: --api-key expected-secret  (CALL_SERVER_TOKEN also set)\n \n== Sibling /api/v1 route WITH Depends(verify_token) ==\n  GET /api/v1/agents [no creds    ] -> HTTP 401\n  GET /api/v1/agents [wrong bearer] -> HTTP 401\n \n== Direct agent-invocation route (the bug) ==\n  POST /agents/owned [no creds      ] -> HTTP 200  {'response': 'EXEC:hello'}\n  POST /agents/owned [wrong bearer  ] -> HTTP 200  {'response': 'EXEC:hello'}\n  POST /agents/owned [wrong x-api-key] -> HTTP 200  {'response': 'EXEC:hello'}\n  POST /agents/owned [empty bearer  ] -> HTTP 200  {'response': 'EXEC:hello'}\n  POST /agents       [no creds      ] -> HTTP 200  {'response': 'EXEC:hi'}\n```\n \nThe auth mechanism works for `/api/v1` (401) and is entirely absent on the direct `/agents` routes (200), despite `--api-key` being configured.\n \n### Equivalent HTTP trigger in a fully installed environment\n \n```bash\npraisonai serve agents --file agents.yaml --host 0.0.0.0 --port 8765 --api-key expected-secret\ncurl -sS -X POST http://TARGET:8765/agents/owned \\\n  -H 'Content-Type: application/json' --data-binary '{\"query\":\"hello\"}'\n# -> 200 {\"response\":\"...\"}  (expected: 401 Unauthorized)\n```\n\n## Impact\n \n- **Direct primitive**: unauthenticated agent invocation despite a configured API key.\n- **Misleading control (aggravating)**: because the CLI advertises `--api-key` as authentication, operators may deliberately expose the service (e.g. `--host 0.0.0.0`, reverse proxy, n8n integration) believing it is protected, increasing the real-world likelihood of exposure.\n- **Downstream**: exposed agents commonly hold LLM provider credentials, RAG/memory, browser/search, MCP, or shell/file tools; the bypass lets an attacker drive those capabilities. Baseline impact is unauthorized LLM cost + access to agent responses.\n\n## Suggested Mitigation\n \n- When `config[\"api_key\"]` is set, build a shared auth dependency and attach it to every agent-invocation / state-changing route in `_create_agents_app()` and `_create_unified_app()` (`dependencies=[Depends(verify)]`).\n- Reuse / unify with the existing `verify_token` so the direct `/agents` routes and the `/api/v1` routes share one mechanism, and wire the CLI `--api-key` into that mechanism (today it feeds nothing; `verify_token` reads `CALL_SERVER_TOKEN`).\n- Use constant-time comparison (`hmac.compare_digest`); `verify_token` currently uses `!=`.\n- Update discovery metadata from `auth_modes=[\"none\"]` to `[\"api-key\",\"bearer\"]` for protected endpoints.\n- Regression tests next to `tests/unit/test_serve_unified.py`: `_create_agents_app({\"api_key\":\"secret\",...})` → `POST /agents/{name}` with no creds / wrong `Authorization` / wrong `X-API-Key` returns `401`; correct key succeeds.\n```python\nimport hmac\nfrom fastapi import Header, HTTPException, Depends\n \ndef _auth_dependency(expected_key: str):\n    async def verify(authorization: str | None = Header(None),\n                     x_api_key: str | None = Header(None, alias=\"X-API-Key\")):\n        token = x_api_key\n        if authorization and authorization.startswith(\"Bearer \"):\n            token = authorization[7:]\n        if not token or not hmac.compare_digest(token, expected_key):\n            raise HTTPException(status_code=401, detail=\"Unauthorized\")\n    return Depends(verify)\n```\n\n## Affected packages\n\n- `praisonai < 4.6.58`\n\n## Remediation\n\nUpgrade to a patched release:\n\n- `praisonai 4.6.58`","depth":"twilight","depthScore":40,"depthScoreParts":{"impact":40.2,"likelihood":0.1,"exploitation":0,"ransomware":0},"changes":[]}