Read 2026-10-23
langchain-ai/langgraph: technical due diligence, read from its repository
What Borehole read in langchain-ai/langgraph's code and history at eb066bd: its licence, security advisories, releases, review practice, tests, dependencies and operations. Every finding cites the file and line it rests on. The full report also covers the team and asks the questions the code cannot answer.
| Dimension | Rating |
|---|---|
| Architecture & Codebase — including AI provenance | strong |
| Engineering Process & SDLC Maturity | strong |
| Product & Engineering Maturity | strong |
| Security & Reliability Posture | adequate |
| Observability & Operations | not assessable |
| Fit with the Acquirer | not assessable |
What the repository says
.github/PULL_REQUEST_TEMPLATE.md:9: states a policy on AI-generated contributions
Architecture & Codebase — including AI provenance
stronglibs/langgraph/langgraph/pregel/main.py is 4,555 lines, 30× this codebase's median file
adequateThe median source file here is 149 lines. 9 files exceed 1,490 lines, the largest being 4,555. Size is not a defect on its own — some problems genuinely live in one place — but these are where merge conflicts, review fatigue and single-owner knowledge concentrate, and they are the first thing that slows an inheriting engineer down.
libs/langgraph/langgraph/pregel/main.py:1 — 4,555 lines (30× median) |
libs/cli/langgraph_cli/deploy.py:1 — 2,941 lines (19× median) |
libs/langgraph/langgraph/pregel/_loop.py:1 — 2,167 lines (14× median) |
libs/prebuilt/langgraph/prebuilt/tool_node.py:1 — 2,065 lines (13× median) |
libs/sdk-py/langgraph_sdk/_async/stream.py:1 — 2,037 lines (13× median) |
Engineering Process & SDLC Maturity
strongAlmost no process scaffolding exists in the repository
adequateOf five ordinary markers — CI workflows, CODEOWNERS, a pull-request template, pre-commit hooks, contribution guidance — 1 is present: CI workflows. Branch protection itself is a GitHub setting rather than a file, so this is a reading of proxies and not of the rules themselves. What it does establish is that the process is not written down anywhere a new engineer would find it.
| 1 of 5 markers present across 698 tracked files |
Tests are retried until they pass
adequate7 tests are individually marked to retry. A retry turns a flaky test green without saying why it failed, so a green build means the tests passed eventually. Ask how many retries a typical run needs, and which tests need them.
libs/checkpoint-sqlite/tests/test_ttl.py:49 — test retry |
libs/checkpoint-sqlite/tests/test_ttl.py:128 — test retry |
Product & Engineering Maturity
strongNo finding.
Security & Reliability Posture
adequate1 credential-shaped string committed to the tree
adequateMatches for JWT appear in tracked files outside test and fixture paths. They stay in the history even after they leave the tree, so what matters is whether each was rotated. The code cannot tell whether any of them is live, and that decides whether this matters: none sits in production configuration, and none is in a format its provider keeps for live keys. Is any of these live? Show that each was rotated after it was committed, and when. 19 other matchs set aside: expired tokens, licence tokens a vendor verifies with a public key, or placeholder text where a key would go. None grants access to anything.
| JWT pattern matched |
1 deserialiser that can run code
adequate1 places read data with a deserialiser that executes what it reads (pickle). That is safe on data the system wrote itself and remote code execution on anything else. Ask where each one's input comes from.
libs/checkpoint/langgraph/checkpoint/serde/jsonplus.py:310 — code-executing deserialiser |
12 locked dependencies have published vulnerabilities, 2 high or critical in what ships
adequate12 of 962 locked dependencies match 21 published advisories in OSV, 12 of them rated high or critical. A lockfile says which version is installed, not whether the flawed code is reachable from this product, so each needs a look. Most are fixed by an upgrade, and the advisory names the version. This lockfile pins development tools alongside what ships and does not say which is which, so some of these may never reach a user.
libs/cli/uv-examples/monorepo/uv.lock — langsmith 0.8.0: GHSA-f4xh-w4cj-qxq8 (high), fixed in 0.8.18 |
libs/cli/uv-examples/simple/uv.lock — langsmith 0.8.5: GHSA-f4xh-w4cj-qxq8 (high), fixed in 0.8.18 |
libs/langgraph/uv.lock — jupyter-server 2.20.0: GHSA-c3mw-737p-c7g2 (high), fixed in 2.21.0; a development dependency |
20 advisories against this project in the last 24 months, 7 high or critical
adequateThe newest was published on 2026-10-21. A project that publishes advisories is telling its users about its fixes, and the count and pace say how much security work it carries. The record is incomplete: NVD: 1 record names langgraph with nothing tying it to this repository (CVE-2025-8709).
| 2026-10-21: GHSA-6hmw-9432-67qq (moderate), fixed in 3.1.2: Namespace labels containing `.` reach other namespaces in Postgres and SQLite stores |
| 2026-10-19: GHSA-8prr-f582-jqq3 (moderate), fixed in 0.4.32: LangGraph CLI prints the registry push token in `langgraph deploy --verbose` output |
| 2026-10-07: GHSA-3fx2-cqw9-xr3c (moderate), fixed in 0.4.6: Unencoded identifiers in LangGraph SDK v3 streaming can redirect authenticated GET requests |
| 2026-09-28: GHSA-4hm6-w6qq-w73v (high), fixed in >=0.15.0: LangGraph JavaScript custom authentication may permit cross-user Store access |
| 2026-09-14: CVE-2026-55235 (moderate): langgraph-api implements the LangGraph API for rapid development and testing. Prior to 0.10.0, langgraph-api permits a run or cron to specify a relative webhook |
Observability & Operations
not assessableNot rated: only 5 of 16 checks could reach a conclusion here; the rest do not apply to this repository or could not read it.
Fit with the Acquirer
not assessableNot rated: structurally not assessable from a repository.
Read more, or read your own
The full report on langchain-ai/langgraph: every dimension, the questions for the room, and the audit log.
Survey a repository: free on public repositories. Private code is read on your own machine by a collector whose source you can read first.