InsForge - Moses Yebei

Abstract

The all-in-one, open-source backend platform for agentic coding. InsForge gives your coding agent database, auth, storage, compute, hosting, and AI gateway to ship full-stack apps end-to-end.

The Problem

Coding agents can generate code, but they lack a backend to deploy it to. Teams end up wiring together databases, auth, storage, and compute manually, or paying for closed platforms that lock the agent in. InsForge positions itself as the open-source layer that gives an agent the same backend primitives a human engineer would have.

What This Does

InsForge is a monorepo of seven self-contained projects, not a single codebase. The core is backend/ (506 files, TypeScript), which implements S3-compatible storage, auth, an AI gateway, email, realtime, and edge functions. packages/ (620 files) holds shared schemas, a UI kit, and a dashboard for managing the platform. frontend/ is a minimal 16-file app, likely a starter or demo.

The repo includes agent integration layers: .agents/, .claude/, .codex/, and .github/copilot-instructions.md define "skills" that let agents read backend state and configure resources. examples/ contains a Python ML experiment tracker. Deployment is via Docker and Kubernetes, with GitHub Actions CI.

How It Is Wired

Execution starts at backend/src/server.ts. The most heavily used path is the S3 gateway: handle in backend/src/api/routes/s3-gateway/commands/abort-multipart-upload.ts reaches 12 functions and is called 45 times by sendS3Error, 22 times by getInstance, and 21 times by getS3Bucket. The request function is called from 130 places, making it the highest-blast-radius change point. withAccessToken (102 callers) and cn (72 callers) follow.

The import graph shows 12 modules in circular dependencies, including packages/dashboard/src/features/database/templates/index.ts and PaymentsSettingsDialog.tsx. The S3 gateway's request.ts and errors.ts are hubs (26 and 24 dependents respectively) — changing them ripples across the codebase. Ten files exceed 1,200 lines, including backend/src/services/auth/auth.service.ts and backend/src/providers/ai/openrouter.provider.ts.

Outbound effects: 4 functions make network calls, 4 do crypto, 13 touch the database, 4 read/write files, and 1 calls a model. The cloud.provider.ts in backend/src/providers/database/ handles both crypto and network, generating connection strings and signing tokens.

How To Use It

git clone https://github.com/moses-y/InsForge
cd InsForge
npm install
cp .env.example .env
# Configure database, auth providers, and storage credentials in .env
npm run dev

The Dockerfile and deploy/Dockerfile.deno support containerized deployment. The README documents an MCP server and CLI + Skills interface for agent interaction. The backend/tests/local/ directory contains docker-compose files for MinIO and RustFS, implying storage backends are swappable.

Real-World Use

A coding agent uses the MCP server to create a Postgres database, deploy an edge function, and set up an auth provider — all through InsForge's API. The agent reads schema metadata, runs a migration, and stores uploads in an S3 bucket, without a human touching infrastructure.

Code Health & Issues

Static analysis found 143 issues: 32 high, 111 medium, 0 low. Five kinds dominate:

High - Hub modules: backend/src/api/routes/s3-gateway/request.ts has 26 dependents; errors.ts has 24. High churn risk.
High - Oversized files: backend/src/services/auth/auth.service.ts at 1,213 lines; openrouter.provider.ts similarly large.
High - Import cycles: 12 modules in circular dependencies, including dashboard feature files.
High - Deep nesting: 32 files with indentation depth 10, e.g., AuthSettingsMenuDialog.tsx.
High - Duplicated code: 1,401 repeated 6-line blocks across 280 files.

The health audit flags three high-severity items: unpinned GitHub Actions (tag-based, vulnerable to supply-chain attacks), a wildcard CORS origin with credentials in backend/src/server.ts, and CI workflows that never run the 277 test files. Medium issues include missing least-privilege token permissions, no Dependabot, unpinned base images, no dependency scanning, and 18MB of video files in docs/images/. Committed secrets were detected.

The Bottom Line

InsForge is a serious, feature-complete backend platform with real depth — S3 gateway, auth, AI providers, and agent tooling. The architecture is sound but carries technical debt: hub modules, deep nesting, and CI that doesn't test. Teams wanting an open-source, agent-friendly backend should evaluate it; teams needing production stability should fix the CI and CORS issues first.

What the analyser found

Deployment readiness

6/7

Container image
CI pipeline
Lockfile committed
Test suite
README
License
No committed secrets

Composition

1,718 files

TypeScript648
TSX308
Markdown109
SQL70
YAML37
Shell31

ReactTailwindDockerKubernetesVector store

Module dependencies

The 10 most depended-upon modules of 969, from static import analysis. Red outline marks a module in an import cycle.

graph LR n0_pi_routes_s3_gateway_request["request"] n1_api_routes_s3_gateway_errors["errors"] n2_backend_src_utils_errors["errors"] n3_rc_api_routes_s3_gateway_xml["xml"] n4_packages_ui_src_lib_index["index"] n5_packages_ui_src_lib_utils["utils"] n6_roviders_oauth_base_provider["base.provider"] n7_res_database_templates_index["index"] n8_c_services_auth_auth_service["auth.service"] n9_s_payments_razorpay_provider["razorpay.provider"] n1_api_routes_s3_gateway_errors --> n3_rc_api_routes_s3_gateway_xml n0_pi_routes_s3_gateway_request --> n1_api_routes_s3_gateway_errors n4_packages_ui_src_lib_index --> n5_packages_ui_src_lib_utils classDef cyc stroke:#ef4444,stroke-width:2px; class n7_res_database_templates_index cyc;

Findings

LowSecret-shaped paths present; the code-health audit confirms or clears them backend/src/infra/database/migrations/009_add-function-secrets.sql, backend/tests/local/test-secrets.sh, openapi/secrets.yaml

What would stop me shipping this

Ranked by severity × confidence × production reach. Reach is the honest discriminator across a collection that is mostly other people's code: the same finding matters more in something that ships.

high3

medium6

low1

Pin third-party GitHub Actions to a commit SHAhigh6 occurrences

.github/workflows

docker/setup-buildx-action@v4, docker/login-action@v4, docker/metadata-action@v6, docker/build-push-action@v7

A tag can be moved, so the action running with your token and secrets is whatever its owner last pushed; this is how tj-actions/changed-files leaked secrets from thousands of repos.

Fix: Replace each @vN with the 40-character commit SHA, keep # vN as a comment, and let Dependabot bump the SHAs.

Replace the wildcard CORS origin with an explicit allow listhigh

backend/src/server.ts

Access-Control-Allow-Origin', '*' together with credentials enabled

Any page on the internet can call the API with the browser's cookies attached, so a logged-in visitor to an unrelated site performs authenticated requests without knowing, and the wildcard combined with credentials is the exact configuration browsers refuse for that reason.

Fix: List the origins the API actually serves, and never pair a wildcard with credentials.

Make CI invoke the test suite it hashigh277 occurrences

.github/workflows

277 test files, no test command in the 4 workflows read

A green check that never executed an assertion is worse than no check, because reviewers trust it.

Fix: Add a test step to the existing workflow rather than a new workflow.

Declare least-privilege permissions for GITHUB_TOKENmedium2 occurrences

.github/workflows/check-migrations.yml

2 workflow(s) declare no permissions

With no declaration the token inherits the repository default, so any injected step can push commits or mint releases from inside your own CI.

Fix: Add permissions: contents: read at the top of the workflow and widen per job only where needed.

Enable Dependabot or Renovatemedium

6 manifest(s), no update bot configured

Without a bot a published advisory sits unpatched until someone audits by hand, which across 1,322 repositories means never.

Fix: Commit .github/dependabot.yml covering the repo ecosystems plus github-actions.

Pin the container base image by digestmedium6 occurrences

Dockerfile

denoland/deno:alpine-${DENO_VERSION}, node:20-alpine, node:20-alpine

An untagged or mutable base means today's build and last month's contain different libc and a different CVE set, with no record of which shipped.

Fix: Use image:tag@sha256:<digest> and enable Dependabot's docker ecosystem.

Gate pull requests on a dependency vulnerability scanmedium

.github/workflows

no dependency scan in CI

This is the one gate that would catch a known-vulnerable package before it reaches a build, and no repository in the sample had it.

Fix: Add dependency-review-action on pull_request, or osv-scanner on push and a schedule.

Move large binaries to Git LFS or out of the repositorymedium2 occurrences

docs/images/integrations/clerk/token-claims.mp4

2 blobs over 5MB: token-claims.mp4 11.9MB, jwt-template.mp4 6.2MB

One repository here carries twenty blobs over 5MB including a 10.9MB spreadsheet, so every clone and every CI checkout pays for data nobody diffs.

Fix: Track those extensions with LFS, or move datasets to object storage and fetch them in a setup step.

Set persist-credentials: false on checkoutmedium

.github/workflows/frontend-tests.yml

checkout keeps the token, then dependencies are installed

The token stays in .git/config for every later step, so a malicious postinstall script reads a pushable credential without one ever being passed to it.

Fix: Add with: persist-credentials: false, and pass an explicit token only to the step that pushes.

Set timeout-minutes on the workflow jobslow4 occurrences

.github/workflows/build-image.yml

4 workflow(s) declare no job timeout

A wedged step runs to the six-hour platform default, which on a two-hourly schedule means three runs overlap behind it.

Fix: Add timeout-minutes with a realistic bound to each job.

Checked deterministically against the repository tree and a bounded set of its files: committed credentials, unpinned actions and base images, missing lockfiles and update bots, workflows that discard failures, published advisories against the declared dependencies, runtime configuration, licensing and notebook reproducibility. No language model is involved in this section.