Drop in SpendGuard in 30 seconds (base URL tools, one config line)

If your agent stack already uses an OpenAI-compatible client, SpendGuard drops in by changing one base URL. Start the SpendGuard egress proxy at http://localhost:9000/v1, point your tool at it, and every model call flows through SpendGuard’s pre-call budget gate and KMS-signed audit chain before it ever hits the upstream provider. No SDK install, no code change, no rebuild. Find your tool in the matrix below, copy the line your tool’s docs specify, and the next call is governed.

How Pattern 2 works

Each of these tools already supports a custom OpenAI-compatible base URL. SpendGuard runs an OpenAI-compatible endpoint of its own; setting the env var (or one config line) routes the tool’s traffic through SpendGuard before it reaches OpenAI / Anthropic / Bedrock. Pattern 2 contrasts with Pattern 1 (in-process SDK middleware, used for the adapter integrations) and Pattern 3 (egress proxy + CA install, used for closed BYOK CLIs like Claude Code and Codex).

Start the proxy locally (30 seconds)

git clone https://github.com/m24927605/agentic-spendguard.git
cd agentic-spendguard
export OPENAI_API_KEY=sk-...
make demo-up DEMO_MODE=proxy

The egress proxy binds on http://localhost:9000/v1. Point any of the tools below at that URL.

Find your tool

The matrix lists the exact env var name or config key each tool’s maintainer docs specify. The Verified column distinguishes rows we have run end-to-end against the egress proxy (Live), rows whose configuration matches the maintainer’s published docs but we have not ourselves smoke-tested (Spec), and rows the tool’s own changelog documents natively without SpendGuard-specific testing (Vendor-native).

Use the quick filter below to narrow by setting type (env var, config file, or admin UI), search by tool name, and jump straight to the per-tool section.

Filter by setting type

All (14) Env var Config file Admin UI Search by tool name

Showing all 14 tools.

Tool	Vendor	Setting type	What to set	Verified	Jump
LiteLLM (proxy mode)	LiteLLM	Env var	export OPENAI_API_BASE=http://localhost:9000/v1	Spec	Jump ↓
Aider	Aider	Env var	export OPENAI_API_BASE=http://localhost:9000/v1	Spec	Jump ↓
Continue	Continue	Config file	apiBase: http://localhost:9000/v1	Spec	Jump ↓
Cline / Roo Code (BYOK)	Cline	Admin UI	OpenAI Compatible provider, Base URL http://localhost:9000/v1	Spec	Jump ↓
OpenHands (BYOK)	All Hands AI	Admin UI	Settings → LLM → Custom Provider → Endpoint	Spec	Jump ↓
Goose	Block	Env var	export OPENAI_HOST=http://localhost:9000	Spec	Jump ↓
Zed AI	Zed	Config file	api_url = "http://localhost:9000/v1"	Vendor-native	Jump ↓
GitHub Copilot CLI (BYOK)	GitHub	Env var	export COPILOT_PROVIDER_BASE_URL=http://localhost:9000/v1	Vendor-native	Jump ↓
Tabnine Enterprise	Tabnine	Admin UI	Admin → AI Models → Connect Custom LLM → Endpoint URL	Spec	Jump ↓
AnythingLLM	Mintplex Labs	Admin UI	Admin → LLM Configuration → Custom OpenAI-compatible base URL	Spec	Jump ↓
LobeChat	LobeHub	Admin UI	Settings → Language Model → Custom Base URL	Spec	Jump ↓
Cody self-hosted	Sourcegraph	Config file	cody.completions.endpoint in site-config.json	Spec	Jump ↓
Augment (BYOK)	Augment	Admin UI	Settings → BYOK → LLM Custom Endpoint	Spec	Jump ↓
Dify	Dify	Config file	Custom Model Provider plugin manifest: base_url	Spec	Jump ↓

The static matrix below carries the upstream-source links and the row numbering the rest of this page references; the picker above is a quick-filter layer over the same data.

| # | Tool | Provider field | Exact value | Upstream source | Verified | Recipe | |---|------|----------------|-------------|-----------------|----------|--------| | 1 | LiteLLM (proxy mode) | Env var | OPENAI_API_BASE=http://localhost:9000/v1 | docs.litellm.ai | Spec | Open ↗ | | 2 | Aider | Env var | OPENAI_API_BASE=http://localhost:9000/v1 | aider.chat | Spec | Open ↗ | | 3 | Continue | YAML apiBase | apiBase: http://localhost:9000/v1 | docs.continue.dev | Spec | Open ↗ | | 4 | Cline / Roo Code (BYOK) | UI field | Custom OpenAI provider, Base URL http://localhost:9000/v1 | docs.cline.bot | Spec | Open ↗ | | 5 | OpenHands (BYOK) | UI field | LLM custom endpoint http://localhost:9000/v1 | docs.all-hands.dev | Spec | Open ↗ | | 6 | Goose | Env var | OPENAI_HOST=http://localhost:9000 | block.github.io/goose | Spec | Open ↗ | | 7 | Zed AI | TOML key | api_url = "http://localhost:9000/v1" | zed.dev | Vendor-native | Open ↗ | | 8 | GitHub Copilot CLI (BYOK) | Env var | COPILOT_PROVIDER_BASE_URL=http://localhost:9000/v1 | docs.github.com | Vendor-native | Open ↗ | | 9 | Tabnine Enterprise | Admin UI | BYO LLM endpoint http://localhost:9000/v1 | docs.tabnine.com | Spec | Open ↗ | | 10 | AnythingLLM | Admin UI | Generic OpenAI provider, Base URL http://localhost:9000/v1 | docs.anythingllm.com | Live | Open ↗ | | 11 | LobeChat | Env var | OPENAI_PROXY_URL=http://localhost:9000/v1 | lobehub.com | Live | Open ↗ | | 12 | Cody self-hosted Enterprise | Site config | Sourcegraph relay endpoint http://localhost:9000/v1 | sourcegraph.com | Spec | Open ↗ | | 13 | Augment (BYOK) | UI field | LLM custom endpoint http://localhost:9000/v1 | docs.augmentcode.com | Spec | Open ↗ | | 14 | Dify | Plugin manifest | Custom Model Provider plugin → base URL http://localhost:9000/v1 | docs.dify.ai | Spec | Open ↗ | | 15 | CrewAI Studio (via LiteLLM) | Indirect | Use row 1 — point CrewAI at the LiteLLM proxy, point LiteLLM at SpendGuard | docs.litellm.ai | Spec | Open ↗ | | 16 | OpenClaw | Config file | models.providers.spendguard.baseUrl: http://localhost:9000/v1 with api: "openai-completions" | github.com/openclaw/openclaw | Live | Open ↗ |

Already running SpendGuard in production? Swap http://localhost:9000/v1 for your SpendGuard egress proxy URL — a Kubernetes Service URL, a sidecar URL inside the pod, or a hosted SpendGuard URL. The relative path is always /v1. See the Helm deployment guide for the production hostname pattern.

Per-tool setup

Each section below is the H3 anchor the matrix above links to. Copy the code block, set it in your shell or config, and the next call from that tool routes through SpendGuard. The full walkthrough — verification steps and known gotchas — lives in each tool’s recipe page in the sidebar.

LiteLLM

LiteLLM proxy mode reads OPENAI_API_BASE for the upstream endpoint when you set the OpenAI provider with no explicit base URL on a route. Export the var before litellm --config config.yaml (or set it in your config.yaml environment block).

export OPENAI_API_BASE=http://localhost:9000/v1

Source: LiteLLM proxy config settings.

Aider

Aider picks up OPENAI_API_BASE from the environment and routes every chat completion through it. Set the env var in the shell you launch aider from.

export OPENAI_API_BASE=http://localhost:9000/v1

Source: Aider OpenAI-compatible APIs guide.

Continue

Continue reads apiBase from each model entry in config.yaml (under the model’s config block for the OpenAI provider). Edit the YAML and reload the IDE extension.

models:
  - name: SpendGuard-gated GPT-4o
    provider: openai
    model: gpt-4o
    config:
      apiBase: http://localhost:9000/v1

Source: Continue model customization.

Cline / Roo Code (BYOK)

In the Cline or Roo Code extension settings, pick the OpenAI Compatible provider and fill the Base URL field. No env var required — this is a UI-only setting that persists in the workspace config.

Provider:  OpenAI Compatible
Base URL:  http://localhost:9000/v1
API Key:   <any value the proxy will forward upstream>

Source: Cline BYOK guide.

OpenHands (BYOK)

OpenHands’ UI exposes a Custom LLM endpoint field under Settings → LLM → Custom Provider. Enter the SpendGuard egress URL there; OpenHands will use it for every model call in the workspace.

Settings → LLM → Custom Provider → Endpoint
http://localhost:9000/v1

Source: OpenHands custom LLM configs.

Goose

Goose reads OPENAI_HOST (host, not full base URL — Goose appends /v1 itself for the OpenAI provider). Export the variable in the shell you launch goose from.

export OPENAI_HOST=http://localhost:9000

Source: Goose installation guide.

Zed AI

Zed AI configuration is TOML inside settings.json (Zed reads JSON or TOML depending on file extension). Add an api_url to the OpenAI provider block.

[ai.openai]
api_url = "http://localhost:9000/v1"

Source: Zed AI configuration.

GitHub Copilot CLI (BYOK)

The GA-2026-04-07 BYOK release of gh copilot honours COPILOT_PROVIDER_BASE_URL for the OpenAI-compatible provider. Export the var; gh copilot picks it up on the next invocation.

export COPILOT_PROVIDER_BASE_URL=http://localhost:9000/v1

Source: GitHub Copilot CLI BYOK reference.

Tabnine Enterprise

Tabnine Enterprise’s admin UI has a Connect Custom LLM form for self-hosted endpoints. Paste the SpendGuard egress URL into the endpoint field and save; the change rolls out to every Tabnine seat on the tenant.

Admin → AI Models → Connect Custom LLM → Endpoint URL
http://localhost:9000/v1

Source: Tabnine connect custom LLM.

AnythingLLM

AnythingLLM ships a Generic OpenAI provider tile under Settings → LLM Preference. Pick the tile, set the Base URL to your SpendGuard egress endpoint (trailing /v1 required), and any non-empty API Key. The full per-tool walkthrough — four exact fields, Docker / Desktop / Cloud notes, the DEMO_MODE=anythingllm_real smoke — lives in the AnythingLLM recipe.

Settings → LLM Preference → Generic OpenAI → Base URL
http://localhost:9000/v1

Source: AnythingLLM custom OpenAI base URL.

LobeChat

LobeChat honours an OPENAI_PROXY_URL env var that rewrites the OpenAI upstream for every server-side chat at /api/chat/openai. Set it at container boot and every chat routes through SpendGuard with no LobeChat code change. The full per-tool walkthrough — Docker / Vercel / Cloud / Client-mode notes, the DEMO_MODE=lobechat_real smoke — lives in the LobeChat recipe.

docker run -e OPENAI_PROXY_URL=http://localhost:9000/v1 \
           -e ACCESS_CODE=... -p 3210:3210 lobehub/lobe-chat:1.40.0

Source: LobeChat — OPENAI_PROXY_URL env var.

OpenClaw

OpenClaw supports custom provider entries under models.providers. Set the provider baseUrl to the SpendGuard egress endpoint and use api: "openai-completions" for OpenAI-compatible chat traffic. The full per-tool walkthrough — fixture config, local / Docker URLs, DEMO_MODE=openclaw_base_url smoke, and D40b plugin boundary — lives in the OpenClaw recipe.

{
  "models": {
    "providers": {
      "spendguard": {
        "baseUrl": "http://localhost:9000/v1",
        "api": "openai-completions"
      }
    }
  }
}

Source: OpenClaw model providers.

Cody self-hosted Enterprise

Sourcegraph Cody self-hosted uses a site-config block for the completions relay endpoint. Edit site-config.json to point the relay at SpendGuard.

{
  "cody.completions.endpoint": "http://localhost:9000/v1"
}

Source: Cody self-hosted install guide.

Augment (BYOK)

Augment’s BYOK flow has an LLM custom endpoint field on the workspace settings page. Paste the SpendGuard URL and save; Augment uses it for every model call.

Settings → BYOK → LLM Custom Endpoint
http://localhost:9000/v1

Source: Augment BYOK setup.

Dify

Dify wires custom OpenAI-compatible endpoints through its Model Provider Plugin mechanism. The plugin manifest declares the base URL; load it into your Dify workspace and pick it as the default provider.

type: model-provider
provider: openai-compatible
base_url: http://localhost:9000/v1

Source: Dify model provider plugin.

What works today

The matrix above is the drop-in surface: tools whose maintainer docs publicly document a custom OpenAI-compatible base URL on 2026-06-06. The Verified column reflects SpendGuard’s own claim per row; for the runtime-verified client SDKs (openai-python, LangChain ChatOpenAI, LangGraph, openai-agents shorthand, streaming) see the README’s What works today table.

What to do next

• Provision a real budget on Kubernetes
• Open the SpendGuard dashboard
• Cover the rest of your stack (Pattern 3 egress proxy)
• Wire a framework SDK (Pattern 1)
• How SpendGuard compares to Cloudflare AI Gateway, Databricks Unity, Portkey

The per-tool deep-dive pages linked in the Recipe column above are the SLICE-2 / D33 / D34 walkthroughs (prerequisites, exact config block, verification step, gotchas). At Slice 1 each Open ↗ link resolves to a Coming soon stub; the env var or config line for that tool is in the matrix above and repeated in the per-tool H3 sections on this page.