Skip to content

Conversation

@betterclever
Copy link
Contributor

@betterclever betterclever commented Jan 15, 2026

Summary

Adds tool mode metadata for agent-callable components, allowing components to be exposed as MCP tools.

Changes

✨ Features

  • Add `agentTool` configuration to component UI metadata
  • Add `ToolMetadata` and `ToolInputSchema` types for MCP compatibility
  • Add helper functions:
    • `isAgentCallable()` - check if a component is configured as an agent-callable tool
    • `inferBindingType()` - infer credential vs action input binding
    • `getCredentialInputIds()` / `getActionInputIds()` - filter inputs by binding type
    • `getToolSchema()` - generate JSON Schema for action inputs only
    • `getToolName()` / `getToolDescription()` - get tool metadata
    • `getToolMetadata()` - get complete MCP-compatible tool metadata

🐛 Bug Fixes

  • Fixed P2 bug: `json`/`any` ports were incorrectly forced to `type: 'object'` in MCP tool schemas
    • Now uses Zod's built-in `toJSONSchema()` for accurate type conversion:
      • `z.any()` → `{}` (empty schema = any JSON value) ✅
      • `z.union([...])` → `{ anyOf: [...] }` ✅
      • `z.enum([...])` → `{ type: 'string', enum: [...] }` ✅
      • `z.literal('X')` → `{ type: 'string', const: 'X' }` ✅
      • `z.record(...)` → `{ type: 'object', additionalProperties: {...} }` ✅

🏗️ Refactoring

  • Uses official `@modelcontextprotocol/sdk` types for `Tool` and `ToolInputSchema`
  • Simplified `tool-helpers.ts` by removing redundant helper functions
  • Reuses existing `getActionInputIds()` function instead of duplicating logic

Testing

All component-sdk tests pass:
```
126 pass
7 skip
0 fail
```

Type Definitions

```typescript
import type { Tool } from '@modelcontextprotocol/sdk/types.js';

// Uses MCP SDK's official Tool.inputSchema type
export type ToolInputSchema = Tool['inputSchema'];

export interface ToolMetadata {
name: string;
description: string;
inputSchema: ToolInputSchema;
}
```

@betterclever betterclever force-pushed the eng-95/component-sdk-tool-mode branch from 13abde0 to 7e4d294 Compare January 15, 2026 16:39
Copy link

@chatgpt-codex-connector chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 7e4d294d12

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment on lines 106 to 108
case 'json':
case 'any':
return 'object';

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Do not force json/any ports to object in tool schema

portTypeToJsonSchemaType maps both json and any to the JSON Schema type object, but coercePrimitive in ports.ts accepts any value for these port types (strings, numbers, arrays, etc.). This makes the generated MCP inputSchema reject legitimate non-object inputs (e.g., a JSON array or string) or mislead agents about the expected shape. Consider omitting type for these ports or using a permissive schema so tool calls that pass valid non-object JSON aren’t rejected.

Useful? React with 👍 / 👎.

@betterclever betterclever force-pushed the eng-95/component-sdk-tool-mode branch 2 times, most recently from 4f4cee6 to f09bb4d Compare January 19, 2026 13:00
betterclever added a commit that referenced this pull request Jan 20, 2026
…emas

Previously, json/any ports were incorrectly forced to type: 'object' in
MCP tool schemas. This fix uses Zod's built-in toJSONSchema() method
which correctly handles all types:

- z.any() → {} (empty schema = any JSON value)
- z.union([...]) → { anyOf: [...] }
- z.enum([...]) → { type: 'string', enum: [...] }
- z.literal('X') → { type: 'string', const: 'X' }
- z.record(...) → { type: 'object', additionalProperties: {...} }

This also simplifies the code by removing redundant helper functions
and reusing the existing getActionInputIds() function.

Fixes P2 bug in PR #207

Signed-off-by: betterclever <[email protected]>
@betterclever betterclever force-pushed the eng-95/component-sdk-tool-mode branch from f09bb4d to 96df26b Compare January 20, 2026 11:31
betterclever added a commit that referenced this pull request Jan 20, 2026
Previously, json/any ports were incorrectly forced to type: 'object' in
MCP tool schemas. This fix uses Zod's built-in toJSONSchema() method
which correctly handles all types:

- z.any() → {} (empty schema = any JSON value)
- z.union([...]) → { anyOf: [...] }
- z.enum([...]) → { type: 'string', enum: [...] }
- z.literal('X') → { type: 'string', const: 'X' }
- z.record(...) → { type: 'object', additionalProperties: {...} }

Changes:
- Use @modelcontextprotocol/sdk for official Tool types
- ToolInputSchema now derives from Tool['inputSchema']
- Simplified code by reusing existing getActionInputIds()
- Removed redundant helper functions

Fixes P2 bug in PR #207

Signed-off-by: betterclever <[email protected]>
@betterclever betterclever force-pushed the eng-95/component-sdk-tool-mode branch from 96df26b to f3da236 Compare January 20, 2026 11:45
…ents

ENG-95

- Add PortBindingType and bindingType to ComponentPortMetadata
- Add AgentToolConfig and agentTool to ComponentUiMetadata
- Create tool-helpers.ts with isAgentCallable, getToolSchema, getCredentialInputIds, getActionInputIds, getToolName, getToolDescription, getToolMetadata
- Add comprehensive tests for all helper functions (14 tests)

Signed-off-by: betterclever <[email protected]>
Previously, json/any ports were incorrectly forced to type: 'object' in
MCP tool schemas. This fix uses Zod's built-in toJSONSchema() method
which correctly handles all types:

- z.any() → {} (empty schema = any JSON value)
- z.union([...]) → { anyOf: [...] }
- z.enum([...]) → { type: 'string', enum: [...] }
- z.literal('X') → { type: 'string', const: 'X' }
- z.record(...) → { type: 'object', additionalProperties: {...} }

Changes:
- Use @modelcontextprotocol/sdk for official Tool types
- ToolInputSchema now derives from Tool['inputSchema']
- Simplified code by reusing existing getActionInputIds()
- Removed redundant helper functions

Fixes P2 bug in PR #207

Signed-off-by: betterclever <[email protected]>
@betterclever betterclever force-pushed the eng-95/component-sdk-tool-mode branch from f3da236 to 0660362 Compare January 20, 2026 11:46
ENG-96

- Create ToolRegistryService with Redis-backed storage
- Implement registerComponentTool, registerRemoteMcp, registerLocalMcp
- Implement getToolsForRun, getTool, getToolByName, getToolCredentials
- Implement areAllToolsReady for agent readiness check
- Implement cleanupRun for workflow completion cleanup
- Encrypt credentials using existing SecretsEncryptionService
- Redis key pattern: mcp:run:{runId}:tools (Hash, TTL 1hr)
- Add McpModule to app imports
- Add comprehensive tests (8 tests passing)

Note: Temporal activities (registerToolActivity, waitForToolsActivity, etc.)
will be added in a follow-up as they reside in the worker package.

Signed-off-by: betterclever <[email protected]>
- Update registerRemoteMcp to store authToken as JSON object for consistency
- Add fallback in getToolCredentials to handle legacy raw string tokens
- Add test case for remote MCP credentials

Signed-off-by: betterclever <[email protected]>
- Add port mapping support to DockerRunner
- Refactor mcp-server component to use dynamic runner config and exposed ports
- Fix trace event mapping in E2E tests
- Add container cleanup to mcp-tool-mode E2E test
- Ensure local MCP registration uses actual containerId and endpoint
- Fix workflow definition validation by adding node positions in test

Signed-off-by: betterclever <[email protected]>
… MCP proxying

- Replace deprecated SSEClientTransport with StreamableHTTPClientTransport
- Fix lint errors (trailing whitespace in constructor and emitProgress)
- Gateway currently executes components inline (to be refactored to Temporal)

Signed-off-by: betterclever <[email protected]>
- Add executeToolCallSignal and toolCallCompletedSignal for MCP tool calls
- Add getToolCallResult query for polling tool execution results
- Refactor callComponentTool to signal workflow instead of inline execution
- Add queryWorkflow method to TemporalService
- Tool calls now execute on worker with full Docker/secrets/storage support

Signed-off-by: betterclever <[email protected]>
…ation

- Refactor component tool execution to run on Temporal workers via signals/queries
- Implement validation for workflow run access and organization ownership
- Add comprehensive telemetry: log tool execution (STARTED, COMPLETED, FAILED) to trace repository
- implement robust external MCP proxying with 30s timeouts and exponential backoff retries
- Add support for tool filtering via allowedTools header
- Add E2E test for MCP gateway tool discovery and execution

Signed-off-by: Antigravity <[email protected]>
Signed-off-by: betterclever <[email protected]>
- Extract X-Run-Id and X-Allowed-Tools headers in McpGatewayController
- Pass organizationId and allowedTools to McpGatewayService
- Add basic protocol version validation
- Fix type casting for MCP transport request handling

Signed-off-by: betterclever <[email protected]>
…eway

- Add McpAuthService to manage short-lived, run-bounded session tokens
- Implement McpAuthGuard for RFC 6750 (Bearer) compliance and AuthInfo injection
- Refactor McpGatewayController to use native MCP AuthInfo instead of internal AuthContext
- Add internal endpoint /internal/mcp/generate-token for session token issuance
- Update E2E tests to validate the complete secure handshake and tool execution flow
- Fix type safety issues in MCP transport integration

Signed-off-by: Antigravity <[email protected]>
Signed-off-by: betterclever <[email protected]>
…script harness

- Ensure component 'parameters' are passed through tool registration and execution signals
- Correctly map agent 'arguments' to component 'inputs' in runComponentActivity
- Fix race condition in logic-script harness by ensuring output directory exists before write
- Update E2E gateway test to reflect correct registration and execution pattern
- Clean up debug logs and resolve linting errors across gateway and worker

Signed-off-by: betterclever <[email protected]>
Detailed plan for enabling graph-based tool→agent binding:
- Phase 1: Compiler tracks tool→agent edges
- Phase 2: Runtime passes connectedToolNodeIds to agent
- Phase 3: Agent queries MCP Gateway for tools
- Phase 4: Gateway filters tools by nodeIds
- Phase 5: E2E tests

Linear-issue: ENG-132
Signed-off-by: betterclever <[email protected]>
Phase 1 of ENG-132:
- Added connectedToolNodeIds to WorkflowNodeMetadata (backend/worker)
- Added tools input port to AI agent component
- Included connectedToolNodeIds in RunComponentActivityInput metadata

Linear-issue: ENG-132
Signed-off-by: betterclever <[email protected]>
Phase 2 of ENG-132:
- Modified compiler to collect connectedToolNodeIds from graph edges
- Updated validator to allow multiple edges to 'tools' port
- Virtualized 'tools' output port for nodes in tool mode
- Updated DTO schemas to support tool mode metadata
- Added unit test to verify tool->agent binding

Linear-issue: ENG-132
Signed-off-by: betterclever <[email protected]>
Phase 3 of ENG-132:
- Updated shipsecWorkflowRun to extract connectedToolNodeIds from node metadata
- Included connectedToolNodeIds in activity metadata for agent discovery
- Synchronized worker types for workflow execution

Linear-issue: ENG-132
Signed-off-by: betterclever <[email protected]>
betterclever and others added 27 commits January 24, 2026 19:54
Adds the new OpenCode agent component with configurable 'providerConfig'. Refactors common gateway token logic into a shared 'utils.ts' used by both AI Agent and OpenCode.

Signed-off-by: betterclever <[email protected]>
- Add zai-coding-plan to LLMProviderSchema with apiKey and modelId support
- Fix OpenCode component to use proper model format (provider/modelId)
- Configure Z.AI provider with apiKey in provider.options
- Fix MCP server config to use type: "remote" instead of transport: "http"
- Remove unused env var API key handling in favor of provider config

Co-Authored-By: Claude <[email protected]>
Signed-off-by: betterclever <[email protected]>
- Use sh -c with properly quoted prompt string to handle multi-word prompts
- Escape single quotes in prompt to prevent shell injection
- Add current-state.md documenting investigation and findings
- Add opencode E2E test

Co-Authored-By: Claude <[email protected]>
Signed-off-by: betterclever <[email protected]>
…er script

- Remove --quiet flag (doesn't exist in opencode 1.1.34), use --log-level ERROR
- Use wrapper script approach to handle prompt file reading inside container
- Set entrypoint to /bin/sh to override default opencode entrypoint
- Fix test assertions to check outputSummary.report instead of output.report
- Update current-state.md with resolution details

E2E tests now passing: 2 pass, 0 fail

Co-Authored-By: Claude <[email protected]>
Signed-off-by: betterclever <[email protected]>
…ided

When a custom systemPrompt is provided, the task was not being included
because the {{TASK}} placeholder only exists in the default template.

Now the task is always appended to ensure OpenCode receives the full prompt.

Co-Authored-By: Claude <[email protected]>
Signed-off-by: betterclever <[email protected]>
- Add optional baseUrl and headers properties to zai-coding-plan provider
- Add zai-coding-plan to ModelProvider type in ai-agent.ts
- This fixes TypeScript build errors when using the new provider

Co-Authored-By: Claude <[email protected]>
Signed-off-by: betterclever <[email protected]>
… d.ts files

- Remove emitDeclarationOnly: true from worker/tsconfig.json
- This allows TypeScript to emit both .js and .d.ts files
- Fixes backend typecheck errors when importing from @shipsec/studio-worker/workflows
- The worker still uses source files directly via bun, so .js files don't interfere

Co-Authored-By: Claude <[email protected]>
Signed-off-by: betterclever <[email protected]>
- Rename unused 'error' to '_error' in catch block to satisfy ESLint
- Apply code formatting from linter to opencode.ts

Co-Authored-By: Claude <[email protected]>
Signed-off-by: betterclever <[email protected]>
Signed-off-by: betterclever <[email protected]>
feat: MCP AWS servers and proxy (ENG-103)

Consolidating the work in 1 PR from the stack
…nt-ui

feat: tool mode UI + agent orchestration (ENG-101/102)

Consolidating the work in one PR
feat: OpenCode agent component (ENG-100)

Consolidating the work in final PR
feat: implement tool-mode orchestration (ENG-132)

Consolidating
feat(mcp): implement MCP Gateway for internal and external tools

Consolidating
feat(dsl): implement workflow tool mode handling and MCP server node
feat(mcp): add Tool Registry Service for agent tool credentials
@betterclever
Copy link
Contributor Author

Superseded by consolidated PR #243 (branch: mcp-tool-mode) that contains the full MCP tool‑mode stack with updated description.

@betterclever
Copy link
Contributor Author

Closing in favor of #243 (full MCP tool‑mode stack consolidated).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants