[MCP][Figma] Resources load but use_figma/get_design_context are not injected into Desktop threads (0.144.5)
Nobody has claimed this yet.
Assessment
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Newbie friendliness
- 45/100
Research direction
Start by reproducing the issue with the [mcp_servers.figma] configuration, codex mcp list --json, and the streamable HTTP endpoint. Trace MCP capability discovery from resources/list and resources/templates/list through per-thread or deferred tool registration, then verify that use_figma and get_design_context appear in a new thread or that a specific diagnostic is exposed.
Written by the indexing model from the issue text.
Description
Summary
Codex Desktop can connect to the official remote Figma MCP server and read its resources, resource templates, documentation, and skill files, but the server's callable tools are not injected into the active thread.
In particular, use_figma and get_design_context are completely absent from the thread tool surface. This blocks both Figma write-to-canvas workflows and normal design-to-code workflows.
Environment
- Codex CLI/app runtime:
codex-cli 0.144.5 - Model:
gpt-5.6-solusing the default OpenAI provider (no custom model provider or proxy configured) - OS: macOS 15.6.1 (24G90), arm64
- Figma MCP endpoint:
https://mcp.figma.com/mcp - Transport:
streamable_http
codex mcp list --json reports:
{
"name": "figma",
"enabled": true,
"disabled_reason": null,
"transport": {
"type": "streamable_http",
"url": "https://mcp.figma.com/mcp"
},
"auth_status": "unsupported"
}
Steps to reproduce
-
Configure the official remote Figma MCP server:
[mcp_servers.figma] url = "https://mcp.figma.com/mcp" -
Authenticate Figma and start a new Codex Desktop thread.
-
Ask Codex to use the official Figma MCP
use_figmaandget_design_contexttools. -
Have the thread enumerate/read Figma MCP resources.
Actual behavior
The Figma MCP server is reachable and its resource APIs work:
resources/listreturns Figma documentation and skills.resources/templates/listreturns Figma resource templates.skill://figma/figma-use/SKILL.mdcan be read successfully.skill://figma/figma-design-to-code/SKILL.mdcan be read successfully.- Other Figma documentation resources can be read normally.
However, no callable mcp__figma__use_figma, mcp__figma__get_design_context, or equivalent deferred tools are exposed to the thread. Explicitly naming the tools and loading the required Figma skills does not make them callable.
This appears to fail between MCP capability discovery and registration in the per-thread/deferred tool surface.
Expected behavior
After the Figma MCP server is enabled and reachable, its supported callable tools should be injected into new Codex threads. At minimum, get_design_context should be available for read-only design-to-code work, and use_figma should be available subject to the Figma account's write permissions.
If a tool is unavailable because of server-side entitlement or capability negotiation, Codex should expose a specific diagnostic rather than presenting resources while silently omitting every callable Figma tool.
Additional information
This closely resembles:
- #20009 — OAuth MCP server is authenticated/configured but
auth_statusremainsunsupportedand tools are not imported into threads. - #19425 — tools are discovered by MCP but not exposed to Desktop threads or
tool_search. - #20771 — MCP servers load successfully but their tools are not surfaced to the model (closed as duplicate of #19425).
It differs from #17106: in that report, mcp__figma__use_figma is dispatched and then hangs/cancels. In this reproduction, the Figma tool schema is never exposed to the thread at all.
Figma's official documentation lists Codex as a supported MCP client, and prior Codex/Figma reports show get_design_context in the normal Figma tool catalog. The remote server and skills are therefore present; only the callable-tool injection layer appears missing.
- Dominant language
- Rust
- Stars
- 125k
- Forks
- 19.5k
- Avg merge
- 1m
- Merged PRs (30d)
- 1k
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from openai/codex
-
enhancement remote
Difficulty 2/5 1-3 hours Newbie friendliness 68/100
-
bug CLI windows-os
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
-
macOS sandbox blocks hw.optional.arm64 sysctl, causing Flutter to misdetect Apple Silicon as x64 Openbug CLI sandbox
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
-
bug CLI TUI
Difficulty 2/5 1-3 hours Newbie friendliness 90/100
-
CLI config enhancement skills
Difficulty 2/5 1-3 hours Newbie friendliness 84/100
Similar issues
-
Difficulty 2/5 1-3 hours Newbie friendliness 86/100
kwakseongjae/auto-hwp#319 ·
-
area:cli bug filter-quality good first issue priority:medium
Difficulty 2/5 1-3 hours Newbie friendliness 84/100
-
Difficulty 1/5 Under an hour Newbie friendliness 72/100
bevyengine/bevy#25861 ·
-
comp-datalake
Difficulty 2/5 1-3 hours Newbie friendliness 88/100
ClickHouse/ClickHouse#121222 ·
-
A-linter
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
oxc-project/oxc#26863 ·