Repository navigation
docs: rename LangSmith Deployment to LangSmith Agent Server - #7616
Conversation
Matches the UI rename (langchainplus#62081) and CLI copy (langgraph#9666). Adds a ProductNames Vale rule.
Distinguish the runtime from Agent Server hosting, retain genuine plural UI labels, and preserve published anchors and compatibility identifiers.
|
Updated the rename to LangSmith Agent Server (Agent Server for short). Agent Servers remains only for multiple resources and literal UI collections; hosting/control-plane capabilities use Agent Server hosting, separate from an individual runtime and the full self-hosted LangSmith installation. Regenerated the affected snippets, retained the four shared-notice consumers and all 29 renamed-heading anchors, and preserved backend identifiers, billing/usage labels, legacy CLI aliases, older-version caveats, and the separate JavaScript CLI follow-up. Changed-file Vale, Ruff/codespell, 62 focused tests, egress consistency, the documentation build, repository-filtered anchor/redirect checks, MDX validation, and Rollout gate: Confirm the renamed UI is live in production before CLI releases or docs publication. Production UI rollout has not been verified. This PR remains draft. |
The standalone langgraph-api repository is archived. The spec PRs now come from the langgraph-api-sync-openapi.yml workflow in langchainplus, which copies langgraph-api/api/openapi.json into this repo.
There was a problem hiding this comment.
Victor Moreira (@victorm-lc) LGTM, I added a few follow-up edits for you to confirm:
-
Minor: Renamed the Vale rule to
LangChain.RetiredProductNames. The new name says what the rule checks: product names that are no longer in use. A comment in the rule file says spelling fixes for current names, such as DeepAgents to Deep Agents, belong inTerminology.yml..vale.iniuses the new name. -
Updated
granular-usage.mdxto use Agent Server and turn the rule back on for it. The page now matches the UI after the rename: the sub-tab is Agent Server and its URL is?tab=agent_server. Both are checked againstGranularUsageDashboard.tsxandgranularUsage.tsin langchainplus. The old?tab=deploymentsURL still redirects to the new tab. The renamed heading keeps its old ID,#langsmith-deployment-usage-kind=langsmith_deployments, so existing links still work.view-usage.mdxandscript-running-pg-support-queries.mdxstay exempt because they quote billing and usage-export labels, such as "LangSmith Deployment Runs", that still use the old name. --> I believe this didn't need to be exempted like view-usage.mdx was? -
Correct where the Agent Server OpenAPI spec comes from.
AGENTS.mdand thedocs-editskill now say that PRs updatingagent-server-openapi.jsonare opened by thelanggraph-api-sync-openapi.ymlworkflow in langchainplus. They no longer point to the standalonelanggraph-apirepository, which is archived. I opened langchainplus#63179 to fix this.
Update: Since this review, I added seven more commits:
- Widened
RetiredProductNames. The rule now also flags LangGraph Platform, LangGraph Deployment, and LangGraph Server, and catches the miscapitalized "Langsmith Deployment", which had hidden one use of the old name. Matching stays case-sensitive on "Deployment" and "Platform", so lowercase "LangSmith deployment" (a self-hosted installation) does not trigger it. - Replaced LangGraph Server with Agent Server on 12 pages and two code samples: runtime, tools, and graph API nodes; MCP auth and connections; AG-UI, backends, and CopilotKit; the Deep Agents Code config file, configuration, and CLI reference pages; and Managed Deep Agents tools. The Polly error
Unable to connect to LangGraph serveron the Azure Terraform pages stays as written because it quotes the UI, now in code formatting so the rule skips it. - Replaced the other retired names it caught: "Langsmith Deployment" in
server-a2a.mdx, "LangGraph platform" in Deep Agentsgoing-to-production.mdxandfrontend/sandbox.mdx, "LangGraph Deployment" insandboxes.mdx, and the "(LangGraph Platform)" comment in the AWS Terraform quick reference, to match the GCP and Azure guides. - Updated the Engine pages to the renamed UI labels.
engine-fix-validation.mdxandengine-setup.mdxnow use the Agent Server section, Baseline Agent Server, Agent Server authentication, and Verify fixes with preview Agent Servers. Theengine-setup.mdxwording follows Robin's feedback. - Renamed "Deployment UI metrics" to "Agent Server UI metrics" in
self-hosted-agent-server-metrics.mdxand its card on the self-hosted overview. The heading keeps#deployment-ui-metricsas its ID, so existing links still work. - Lowercased "deployment" in
api-ref-control-plane.mdx, where it names the API resource, not the product.
|
Mintlify preview branch generated: Site preview: https://langchain-5e9cc07a-preview-victor-rename-agent-serv-dc512515.mintlify.site Important Preview links may take a few minutes to start working while the deployment finishes. Changed documentation pages (preview deep links):
Only the top 5 changed markdown files by diff size are listed. |
## Summary agents running on agent servers in langsmith cloud can already call a tunnel url. they just send the agent server's own `LANGSMITH_API_KEY` as `X-Api-Key`, which every agent server already has. the tunnels page didn't list agent servers as a consumer, so this adds them and explains the headers. it uses the new agent server name from #7616. - adds agent servers to the list of places you can use a tunnel url - new "call a tunnel from an agent server" section with a python example - covers `X-Tunnel-Origin-X-Api-Key` / `X-Tunnel-Origin-Cookie` for services that use their own key or cookie - notes that any agent server in the org can call any tunnel in it Part of [LSD-2176](https://linear.app/langchain/issue/LSD-2176/support-tunnels-for-langsmith-deployments-cloud-agent-server-to) ## Test Plan - tested end to end on dev: a deployed agent called a tunnel url with its `LANGSMITH_API_KEY` and got the service's response (200). without the key it got `403 connection access denied`, and with the tunnel disabled `403 connection is disabled`. `X-Tunnel-Origin-X-Api-Key` arrived at the service as `X-Api-Key`, and LangSmith's own key was stripped - the code example matches the code that ran in that test - vale on `src/langsmith/tunnels.mdx`: 0 errors, 0 warnings, 0 suggestions - link targets exist: `/langsmith/deploy-to-cloud`, `/langsmith/control-plane#langsmith-integration` - not run locally: `docs dev` and the broken-links check. relying on the preview and CI Prepared with Claude Code. ## Demo <!-- Drag a screen recording or screenshot here --> --------- Co-authored-by: Sean Angiolillo <sean.angiolillo@langchain.dev>
… from Agent Server (#7687) ## Why Follow-up to #7616, which renamed LangSmith Deployment to LangSmith Agent Server across the docs but deliberately kept every file name and URL slug. That left links that read `[Agent Server](/langsmith/deployment)`, and the product team asked for `/deployment` to change (DOC-1922). Renaming the files turned up a bigger problem than stale slugs. The obvious new home for `deployment.mdx`, `agent-server.mdx`, already existed. With `agent-server-overview.mdx`, the docs had **three pages titled some form of "Agent Server", all with the sidebar title "Overview"**: | Page | Title | Nav location | |------|-------|--------------| | `deployment.mdx` | LangSmith Agent Server | Deploy > Get started | | `agent-server-overview.mdx` | Agent Server | Agent Server tab landing page | | `agent-server.mdx` | Agent Server | Agent Server > Capabilities | After the rename, a reader could no longer tell from titles alone where to go to *deploy an agent* and where to go to *learn the Agent Server runtime*. Moving files would have hidden this problem without fixing it. ## The main change: separate the task from the product This PR gives each question its own landing page: - **`/langsmith/deploy-overview`, "Deploy agents"** (formerly `/langsmith/deployment`). This page answers the task question: I have an agent, so where and how do I run it in production? It covers choosing a framework, choosing a hosting environment (Cloud, hybrid, self-hosted, standalone), pricing, and what to do after deploying. It stays the first page of the Deploy > Get started tab. - **`/langsmith/agent-server`, "LangSmith Agent Server"**. This page answers the product question: what is the runtime, and how does it work? `agent-server-overview.mdx` is merged into it, and it becomes the landing page of the Agent Server tab. It covers assistants, threads, runs, persistence, the task queue, runtime architecture, auth, customization, and tutorials. Each page links to the other near the top, so a reader who lands on the wrong one gets redirected in one click. ### Why `deploy-overview` and not an `agent-server-*` slug This is the change most likely to draw debate, so here is the reasoning: - **The page is about an action, not a product.** Its content is a decision guide: which framework and which environment. That is true whichever product name it carries. A task-based title ("Deploy agents") also matches the menu item it sits under (Deploy) and the way readers search ("deploy langgraph agent", not "agent server"). - **It frees `agent-server` to mean one thing.** If the deploy page took an `agent-server-*` slug, the docs would still have two overlapping Agent Server landing pages, which is the problem this PR is trying to fix. Now the product name points at the product page. - **It stays a single page.** Deploy pages for specific environments (`deploy-to-cloud`, `deploy-with-control-plane`, `deploy-standalone-server`) already use verb-led slugs. `deploy-overview` fits with them. The tradeoff is that "LangSmith Agent Server" no longer appears in the deploy page's title. The page still opens by naming and linking Agent Server as the runtime that agents deploy to, and the description keeps the product name for search. ## Other changes - **Product-named slugs renamed** with `docs mv`, which rewrote every inbound link and nav entry, and a redirect added for each old URL: - `deployment` → `deploy-overview` - `deployment-revisions` → `agent-server-revisions` (title now "Agent Server revisions") - `manage-deployment` → `manage-agent-server` - `monitor-deployment` → `monitor-agent-server` - `enable-self-hosted-deployment` → `enable-self-hosted-agent-server` - **`agent-server-overview` merged into `agent-server`**, with a redirect. The merged page adds what the old overview had and the runtime page lacked: streaming, human-in-the-loop, double-texting, webhooks, MCP/A2A, dependency setup, auth and customization cards, and tutorials. It also adds new, explicit coverage of who provisions PostgreSQL and Redis on Cloud, self-hosted, and standalone setups. - **`deployment-quickstart-da` deleted.** It was a `noindex` `url:` stub that was not in the nav and had no inbound links. A redirect to the Managed Deep Agents quickstart replaces it. - **Existing redirects retargeted** so nothing chains: `/langgraph-platform/index`, `/langsmith/deployments`, `/langsmith/develop-agents-overview`, and `/langsmith/deploy-self-hosted-full-platform` now point to the new slugs. Two redirects that pointed at `/langsmith/deployment#agent-server` now go to `/langsmith/agent-server`. - **Inbound link text fixed** across LangSmith, OSS, and snippet pages, where links to the deploy page read "Agent Server" but now point at a page titled "Deploy agents" (and vice versa). ### Intentionally left alone - **`deployment-quickstart`**: its title is "Deploy your app to cloud", so "deployment" describes the task, not the product. - **`smith-deployments-sdk`**: a `url:` stub pointing at `reference.langchain.com/.../deployment/sdk/`. That URL lives outside this repo, so changing it is out of scope. - **The 12 remaining `{#langsmith-deployment-…}` heading IDs** (billing, usage, Terraform, and similar pages): they protect external `#fragment` links and cost nothing to keep. The one on the old deploy page's environments heading was dropped, because the heading changed and nothing in the repo links to it. ## Review notes - **Information architecture**: please weigh in on the `deploy-overview` / "Deploy agents" naming above in particular. - **Product links (follow-up, product team)**: two links in the LangSmith product still point at `/langsmith/deployment` and `/langsmith/deployment-revisions`. The new redirects cover both, but the product team should update them to the new slugs. [I could probably also point Robin at this...] Linear: DOC-1922 --------- Co-authored-by: robin-agent[bot] <215916821+robin-agent[bot]@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Why
Rename LangSmith Deployment to LangSmith Agent Server. Use Agent Server as the short product/runtime name, and Agent Servers only for multiple resources or literal UI collection labels.
Changes
--agent-server-idand--agent-server-typewhere implemented. Preserve hidden legacy aliases, older-version help checks, pinned agent-binding requirements, source-dependent update eligibility, and the separate JavaScript LangGraph CLI follow-up.LangChain.ProductNamesVale replacement to singular. Preserve billing/usage labels, compatibility identifiers, commands, environment variables, API fields, machine-readable output, and generated changelog history.Review notes
Related
Prepared with Claude Code.