feat: support StreamableHTTPClientTransport #64

leo237 · 2025-04-22T10:07:10Z

Description

Adds support for Streamable HTTP. Related to #59

Fixes # (issue)

Type of change

Bug fix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (fix or feature that would cause existing functionality to not work as expected)
Documentation update
Refactoring (no functional changes)

How Has This Been Tested?

Tests are added
Manually tested with a Streamable HTTP server

Checklist:

My code follows the style guidelines of this project
I have performed a self-review of my own code
I have commented my code, particularly in hard-to-understand areas
I have made corresponding changes to the documentation
My changes generate no new warnings
I have added tests that prove my fix is effective or that my feature works
New and existing unit tests pass locally with my changes

benjamincburns · 2025-05-06T00:35:06Z

Per the MCP spec, the streamable HTTP transport is meant to replace the SSE transport. Per their backward compatibility guide, clients should attempt to connect to URLs as though they are streamable HTTP first, and then fall back to SSE.

Doing this would fix another problem. Prior to this change the transport type can be inferred from the structure of the config entry. That is, the transport key of the config was optional, and should remain so. Supporting this fallback behavior would allow for that to continue.

benjamincburns · 2025-05-06T17:40:17Z

I pushed some updates for the above feedback, but I only had time last night to rework some of the tests. IIRC there are 3 tests still failing. We should also make updates to the README & examples to drop the transport field (it's optional and not necessary in 99% of cases).

benjamincburns · 2025-05-12T18:04:42Z

The build is green, however I added a new example to capture the Streamable HTTPs w/ fallback to SSE behavior, and it's not running successfully on the default (Streamable HTTP) case. Seems to work fine for the SSE case, however.

Dandelion-F · 2025-05-13T15:45:37Z

When would this PR be merged? I'm in an urgent need of it. 🙏

benjamincburns · 2025-05-13T18:25:17Z

OK, I think this is ready to go, but because I've mostly rewritten the original PR I need to request an internal review from another colleague.

.gitignore

benjamincburns · 2025-05-13T18:28:11Z

README.md

+| `prefixToolNameWithServerName` | boolean | `true`  | If true, prefixes all tool names with the server name (e.g., `serverName__toolName`) |
+| `additionalToolNamePrefix`     | string  | `mcp`   | Additional prefix to add to tool names (e.g., `prefix__serverName__toolName`)        |


These defaults weren't actually changed in this PR (see checks in tests to verify) - just updated in README and config schema to reflect the actual defaults

benjamincburns · 2025-05-13T18:29:22Z

__tests__/client.comprehensive.test.ts

+      const conf = client.config;
+      expect(conf.additionalToolNamePrefix).toBe("mcp");
+      expect(conf.prefixToolNameWithServerName).toBe(true);


Added this during test troubleshooting because the defaults reverted due to me accidentally setting default(...).optional() in the zod schema updates (needs to be .optional().default(...) else the defaults aren't applied during parsing).

benjamincburns · 2025-05-13T18:32:29Z

examples/calculator_server_shttp_sse.ts

@@ -0,0 +1,173 @@
+import express from "express";


Added this file for manual testing & for use via the added example because we don't yet have an integration suite.

benjamincburns · 2025-05-13T18:48:59Z

src/client.ts

+    if (code == null) {
+      const m = streamableError.message.match(/\(HTTP (\d\d\d)\)/);
+      if (m && m.length > 1) {
+        code = parseInt(m[1], 10);
+      }
+    }


This is probably the grossest part - I needed this during early testing because for some reason the code property of the error wasn't being set. It appears that it is being set now, but I decided to leave it in for reliability.

benjamincburns · 2025-05-13T19:25:04Z

When would this PR be merged? I'm in an urgent need of it. 🙏

@Dandelion-F will depend on code review - if all goes well it could be merged/released within a few hours.

vbarda · 2025-05-13T19:33:38Z

__tests__/client.basic.test.ts

          url: "http://localhost:8000/sse",
        },
+        server3: {
+          transport: "streamable",


(nit) why "streamable" and not "streamable-http" or "streamable_http"?

I didn't decide the key - honestly I think http is better since this option should be used for all HTTP endpoints (including SSE). Will change to that.

Changed - will likely need to add strings to this union as other popular client configs come out and decide they want to use some other key, however.

benjamincburns · 2025-05-13T20:01:17Z

(code was approved internally by @vbarda)

Thanks for kicking this off @leo237!

benjamincburns changed the title ~~Feature/streamable http~~ feat: support StreamableHttpTransport May 6, 2025

benjamincburns changed the title ~~feat: support StreamableHttpTransport~~ feat: support StreamableHTTPTransport May 6, 2025

benjamincburns changed the title ~~feat: support StreamableHTTPTransport~~ feat: support StreamableHTTPClientTransport May 6, 2025

Abhilash Panigrahi and others added 8 commits May 12, 2025 09:08

Added support for basic streamable http adapter

0ca7648

added tests

b265abf

updated example

07ff66b

Updated readme

e88ef0c

fixes formatting issues

05f4b67

Use Streamable HTTP for all URLs first, fallback to SSE

ee26a0f

partial test rework

c193c20

fix tests, add example (not working yet)

b33a2de

benjamincburns force-pushed the feature/streamable-http branch from 7b94c69 to b33a2de Compare May 12, 2025 17:53

benjamincburns reviewed May 13, 2025

View reviewed changes

.gitignore Show resolved Hide resolved

benjamincburns force-pushed the feature/streamable-http branch from 2002b2e to 9be061b Compare May 13, 2025 18:27

fixes, add Streaming HTTP example w/ SSE fallback, update REAMDE

0f42f28

benjamincburns force-pushed the feature/streamable-http branch from 9be061b to 0f42f28 Compare May 13, 2025 18:35

bump to latest MCP sdk

ce96b23

benjamincburns reviewed May 13, 2025

View reviewed changes

minor tweak

84f0f70

benjamincburns force-pushed the feature/streamable-http branch from 6b352ee to 84f0f70 Compare May 13, 2025 19:28

vbarda reviewed May 13, 2025

View reviewed changes

use "http" as transport ID instead of "streamable"

972158c

benjamincburns merged commit 68f130e into langchain-ai:main May 13, 2025

		\| `prefixToolNameWithServerName` \| boolean \| `true` \| If true, prefixes all tool names with the server name (e.g., `serverName__toolName`) \|
		\| `additionalToolNamePrefix` \| string \| `mcp` \| Additional prefix to add to tool names (e.g., `prefix__serverName__toolName`) \|

feat: support StreamableHTTPClientTransport #64

feat: support StreamableHTTPClientTransport #64

Uh oh!

Conversation

leo237 commented Apr 22, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Type of change

How Has This Been Tested?

Checklist:

Uh oh!

benjamincburns commented May 6, 2025

Uh oh!

benjamincburns commented May 6, 2025

Uh oh!

benjamincburns commented May 12, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Dandelion-F commented May 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

benjamincburns commented May 13, 2025

Uh oh!

Uh oh!

benjamincburns May 13, 2025

Choose a reason for hiding this comment

Uh oh!

benjamincburns May 13, 2025

Choose a reason for hiding this comment

Uh oh!

benjamincburns May 13, 2025

Choose a reason for hiding this comment

Uh oh!

benjamincburns May 13, 2025

Choose a reason for hiding this comment

Uh oh!

benjamincburns commented May 13, 2025

Uh oh!

vbarda May 13, 2025

Choose a reason for hiding this comment

Uh oh!

benjamincburns May 13, 2025

Choose a reason for hiding this comment

Uh oh!

benjamincburns May 13, 2025

Choose a reason for hiding this comment

Uh oh!

benjamincburns commented May 13, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

leo237 commented Apr 22, 2025 •

edited

Loading

benjamincburns commented May 12, 2025 •

edited

Loading

Dandelion-F commented May 13, 2025 •

edited

Loading