feat: MCP first MVP #97
Open
+263
−0
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
WalkthroughAdds an MCP server exposing evaluate and check tools for Numscript, a hidden Cobra subcommand to run it, registers the subcommand with the root command, and adds a helper to convert diagnostic severities to plain strings. Changes
Sequence Diagram(s)sequenceDiagram
autonumber
actor User
participant CLI as CLI (cobra root)
participant MCP as mcpCmd
participant Server as MCP Server
participant Tools as Tools (evaluate/check)
participant Engine as Numscript Engine
participant Anal as Analyzer
User->>CLI: run app mcp
CLI->>MCP: Execute mcpCmd
MCP->>Server: RunServer()
Server-->>User: serve over stdio (MCP)
rect rgb(230,245,255)
note right of User: evaluate flow
User->>Server: tool:evaluate {script, balances, vars}
Server->>Tools: invoke evaluate handler
Tools->>Tools: parse balances & vars
Tools->>Engine: interpreter.RunProgram(...)
alt success
Engine-->>Tools: result
Tools-->>Server: ToolResultJSON
Server-->>User: success response
else error
Engine-->>Tools: error
Tools-->>Server: ToolResultError
Server-->>User: error response
end
end
rect rgb(240,255,240)
note right of User: check flow
User->>Server: tool:check {script}
Server->>Tools: invoke check handler
Tools->>Anal: parse + analyze(script)
Anal-->>Tools: diagnostics [kind,severity,span]
Tools-->>Server: ToolResultJSON {errors:[...]}
Server-->>User: response
end
Estimated code review effort🎯 4 (Complex) | ⏱️ ~55 minutes Poem
Pre-merge checks and finishing touches❌ Failed checks (1 warning)
✅ Passed checks (2 passed)
✨ Finishing touches
🧪 Generate unit tests (beta)
📜 Recent review detailsConfiguration used: CodeRabbit UI Review profile: CHILL Plan: Pro Disabled knowledge base sources:
📒 Files selected for processing (1)
🧰 Additional context used🧬 Code graph analysis (1)internal/mcp_impl/handlers.go (3)
🔇 Additional comments (1)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
ascandone
commented
Oct 7, 2025
| server.WithInstructions(` | ||
| You're a numscript expert AI assistant. Numscript is a DSL that allows modelling finantial transaction in an easy and declarative way. Numscript scripts alwasy terminate. | ||
| `), | ||
| // TODO add prompt |
There was a problem hiding this comment.
we can add a more complete prompt in future releases
ascandone
commented
Oct 7, 2025
| interpreter.StaticStore{ | ||
| Balances: balances, | ||
| }, | ||
| map[string]struct{}{}, |
There was a problem hiding this comment.
we'll add feature flags in future releases
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #97 +/- ##
==========================================
- Coverage 70.89% 68.49% -2.41%
==========================================
Files 42 44 +2
Lines 4821 4993 +172
==========================================
+ Hits 3418 3420 +2
- Misses 1247 1418 +171
+ Partials 156 155 -1 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 4
🧹 Nitpick comments (1)
internal/cmd/mcp.go (1)
8-22: Set Silence flags on the command itself (optional).Avoids duplicate error/usage output without toggling inside RunE.
var mcpCmd = &cobra.Command{ Use: "mcp", Short: "Run the mcp server", Hidden: true, + SilenceErrors: true, + SilenceUsage: true, RunE: func(cmd *cobra.Command, args []string) error { err := mcp_impl.RunServer() if err != nil { - cmd.SilenceErrors = true - cmd.SilenceUsage = true return err } return nil }, }
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
Disabled knowledge base sources:
- Linear integration is disabled by default for public repositories
You can enable these sources in your CodeRabbit configuration.
⛔ Files ignored due to path filters (2)
go.modis excluded by!**/*.modgo.sumis excluded by!**/*.sum,!**/*.sum
📒 Files selected for processing (4)
internal/analysis/diagnostic_kind.go(1 hunks)internal/cmd/mcp.go(1 hunks)internal/cmd/root.go(1 hunks)internal/mcp_impl/handlers.go(1 hunks)
🧰 Additional context used
🧬 Code graph analysis (3)
internal/analysis/diagnostic_kind.go (1)
internal/utils/utils.go (1)
NonExhaustiveMatchPanic(8-10)
internal/mcp_impl/handlers.go (3)
internal/interpreter/interpreter.go (1)
RunProgram(228-285)internal/analysis/check.go (1)
CheckSource(296-304)internal/analysis/diagnostic_kind.go (2)
SeverityToString(39-52)Severity(12-12)
internal/cmd/mcp.go (1)
internal/mcp_impl/handlers.go (1)
RunServer(175-196)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (2)
- GitHub Check: Tests
- GitHub Check: Dirty
🔇 Additional comments (3)
internal/analysis/diagnostic_kind.go (1)
39-52: Helper looks correct and consistent with existing mapping.internal/cmd/root.go (1)
27-27: Wiring the hidden mcp subcommand looks good.internal/mcp_impl/handlers.go (1)
12-14: Confirmed valid mcp-go version pin
go.mod is pinned to v0.41.1, which matches an existing release.
Comment on lines
+33
to
+41
| for asset, amountRaw := range assets { | ||
| amount, ok := amountRaw.(float64) | ||
| if !ok { | ||
| return interpreter.Balances{}, mcp.NewToolResultError(fmt.Sprintf("Expected float for amount: %v", amountRaw)) | ||
| } | ||
|
|
||
| n, _ := big.NewFloat(amount).Int(new(big.Int)) | ||
| iBalances[account][asset] = n | ||
| } |
There was a problem hiding this comment.
Amounts parsing risks precision loss and silently truncates fractions. Enforce integer semantics and accept safe types.
- Converting via big.NewFloat(amount).Int ignores fractional parts and float64 precision limits (money correctness risk).
- Validate integer-ness, reject NaN/Inf, and optionally accept string integers.
Apply:
- for asset, amountRaw := range assets {
- amount, ok := amountRaw.(float64)
- if !ok {
- return interpreter.Balances{}, mcp.NewToolResultError(fmt.Sprintf("Expected float for amount: %v", amountRaw))
- }
-
- n, _ := big.NewFloat(amount).Int(new(big.Int))
- iBalances[account][asset] = n
- }
+ for asset, amountRaw := range assets {
+ var n *big.Int
+ switch v := amountRaw.(type) {
+ case float64:
+ if math.IsNaN(v) || math.IsInf(v, 0) {
+ return interpreter.Balances{}, mcp.NewToolResultError(fmt.Sprintf("Invalid number for amount at %s/%s: %v", account, asset, v))
+ }
+ if v != math.Trunc(v) {
+ return interpreter.Balances{}, mcp.NewToolResultError(fmt.Sprintf("Amount must be an integer at %s/%s, got: %v", account, asset, v))
+ }
+ if v > float64(math.MaxInt64) || v < float64(math.MinInt64) {
+ return interpreter.Balances{}, mcp.NewToolResultError(fmt.Sprintf("Amount out of 64-bit range at %s/%s: %v", account, asset, v))
+ }
+ n = big.NewInt(int64(v))
+ case string:
+ n = new(big.Int)
+ if _, ok := n.SetString(v, 10); !ok {
+ return interpreter.Balances{}, mcp.NewToolResultError(fmt.Sprintf("Invalid integer string for amount at %s/%s: %q", account, asset, v))
+ }
+ default:
+ return interpreter.Balances{}, mcp.NewToolResultError(fmt.Sprintf("Expected integer (number or string) for amount at %s/%s, got: %T", account, asset, amountRaw))
+ }
+ iBalances[account][asset] = n
+ }Additionally, add import:
@@
-import (
+import (
"context"
"fmt"
+ "math"
"math/big"
"strings"Committable suggestion skipped: line range outside the PR's diff.
Comment on lines
+55
to
+58
| value, ok := rawValue.(string) | ||
| if !ok { | ||
| return map[string]string{}, mcp.NewToolResultError(fmt.Sprintf("Expected stringified var, got: %v", key)) | ||
| } |
There was a problem hiding this comment.
Fix error message: report the offending value/type, not the key.
- value, ok := rawValue.(string)
- if !ok {
- return map[string]string{}, mcp.NewToolResultError(fmt.Sprintf("Expected stringified var, got: %v", key))
- }
+ value, ok := rawValue.(string)
+ if !ok {
+ return map[string]string{}, mcp.NewToolResultError(fmt.Sprintf("Variable %q must be a string, got: %T", key, rawValue))
+ }🤖 Prompt for AI Agents
In internal/mcp_impl/handlers.go around lines 55 to 58, the error message
currently reports the key when the cast to string fails; change it to report the
offending value and its type instead. Replace the fmt.Sprintf argument so it
includes rawValue and reflect.TypeOf(rawValue) (or use %T) in the message, e.g.
"Expected stringified var, got value=%v (type=%T)", rawValue, rawValue, and
return that in mcp.NewToolResultError.
Comment on lines
98
to
137
| s.AddTool(tool, func(ctx context.Context, request mcp.CallToolRequest) (*mcp.CallToolResult, error) { | ||
| script, err := request.RequireString("script") | ||
| if err != nil { | ||
| return mcp.NewToolResultError(err.Error()), nil | ||
| } | ||
|
|
||
| parsed := parser.Parse(script) | ||
| if len(parsed.Errors) != 0 { | ||
| // TODO return all errors | ||
| out := make([]string, len(parsed.Errors)) | ||
| for index, err := range parsed.Errors { | ||
| out[index] = err.Msg | ||
| } | ||
| mcp.NewToolResultError(strings.Join(out, ", ")) | ||
| } | ||
|
|
||
| balances, mcpErr := parseBalancesJson(request.GetArguments()["balances"]) | ||
| if mcpErr != nil { | ||
| return mcpErr, nil | ||
| } | ||
|
|
||
| vars, mcpErr := parseVarsJson(request.GetArguments()["vars"]) | ||
| if mcpErr != nil { | ||
| return mcpErr, nil | ||
| } | ||
|
|
||
| out, iErr := interpreter.RunProgram( | ||
| context.Background(), | ||
| parsed.Value, | ||
| vars, | ||
| interpreter.StaticStore{ | ||
| Balances: balances, | ||
| }, | ||
| map[string]struct{}{}, | ||
| ) | ||
| if iErr != nil { | ||
| mcp.NewToolResultError(iErr.Error()) | ||
| } | ||
| return mcp.NewToolResultJSON(*out) | ||
| }) |
There was a problem hiding this comment.
Evaluate handler: missing returns on error paths and potential nil deref; also use ctx and avoid variable shadowing.
- On parse errors, you create a ToolResultError but don’t return it.
- On interpreter errors, same issue; returning JSON with a nil deref is possible.
- Use the provided ctx instead of context.Background.
- Avoid reusing the name out.
Apply:
@@
- parsed := parser.Parse(script)
- if len(parsed.Errors) != 0 {
- // TODO return all errors
- out := make([]string, len(parsed.Errors))
- for index, err := range parsed.Errors {
- out[index] = err.Msg
- }
- mcp.NewToolResultError(strings.Join(out, ", "))
- }
+ parsed := parser.Parse(script)
+ if len(parsed.Errors) != 0 {
+ msgs := make([]string, len(parsed.Errors))
+ for i, pe := range parsed.Errors {
+ msgs[i] = pe.Msg
+ }
+ return mcp.NewToolResultError(strings.Join(msgs, ", ")), nil
+ }
@@
- out, iErr := interpreter.RunProgram(
- context.Background(),
+ execRes, iErr := interpreter.RunProgram(
+ ctx,
parsed.Value,
vars,
interpreter.StaticStore{
Balances: balances,
},
map[string]struct{}{},
)
- if iErr != nil {
- mcp.NewToolResultError(iErr.Error())
- }
- return mcp.NewToolResultJSON(*out)
+ if iErr != nil {
+ return mcp.NewToolResultError(iErr.Error()), nil
+ }
+ return mcp.NewToolResultJSON(execRes)🤖 Prompt for AI Agents
internal/mcp_impl/handlers.go around lines 98-137: the handler builds
ToolResultErrors but doesn’t return them, may dereference a nil interpreter
result, uses context.Background instead of the provided ctx, and reuses variable
names (out, mcpErr) causing shadowing. Fix by returning the error results
immediately (e.g. return mcp.NewToolResultError(strings.Join(...)), nil) on
parse failures and interpreter errors, check the interpreter output before
dereferencing so you don’t return mcp.NewToolResultJSON on a nil pointer,
replace context.Background() with the incoming ctx, and rename shadowed
variables (e.g. parsedErrs/outResult/parseErr) to avoid reuse. Ensure every
branch returns a valid (*mcp.CallToolResult, error).
Comment on lines
+182
to
+184
| server.WithInstructions(` | ||
| You're a numscript expert AI assistant. Numscript is a DSL that allows modelling finantial transaction in an easy and declarative way. Numscript scripts alwasy terminate. | ||
| `), |
There was a problem hiding this comment.
Typos in instructions string.
Correct spelling/grammar to avoid user-facing nits:
- server.WithInstructions(`
- You're a numscript expert AI assistant. Numscript is a DSL that allows modelling finantial transaction in an easy and declarative way. Numscript scripts alwasy terminate.
- `),
+ server.WithInstructions(`
+ You're a Numscript expert AI assistant. Numscript is a DSL that allows modeling financial transactions in an easy and declarative way. Numscript scripts always terminate.
+ `),📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| server.WithInstructions(` | |
| You're a numscript expert AI assistant. Numscript is a DSL that allows modelling finantial transaction in an easy and declarative way. Numscript scripts alwasy terminate. | |
| `), | |
| server.WithInstructions(` | |
| You're a Numscript expert AI assistant. Numscript is a DSL that allows modeling financial transactions in an easy and declarative way. Numscript scripts always terminate. | |
| `), |
🤖 Prompt for AI Agents
In internal/mcp_impl/handlers.go around lines 182 to 184, the instruction string
contains typos and grammar errors; replace the existing text with a corrected
sentence such as: "You are a Numscript expert AI assistant. Numscript is a DSL
that models financial transactions in an easy, declarative way. Numscript
scripts always terminate." Ensure spelling (Numscript, financial, always) and
punctuation are fixed and the string remains wrapped as a single
server.WithInstructions(...) argument.
Dav-14
reviewed
Oct 7, 2025
internal/mcp_impl/handlers.go
Outdated
| } | ||
|
|
||
| out, iErr := interpreter.RunProgram( | ||
| context.Background(), |
There was a problem hiding this comment.
You might want to use the proper context or a derived one ?
There was a problem hiding this comment.
This one should be ok: we are only doing in-memory operations, and fetching data in a go struct
(We need context because the implementor of the interface might be sql queries )
There was a problem hiding this comment.
David is right.
Always use the correct context.
There is no good reason to not use it.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Just an initial working MPV of an MCP integration for numscript. There are endless improvement that can be made, by improving current prompts and adding many other tools, but for now that's 80% of the value we can get out of it, with not so much complexity