# Router.com (Ramp Router) (slim)

> Router.com (Ramp Router) is Ramp's LLM gateway. One API key reaches OpenAI, Anthropic, xAI and open-weight models through OpenAI Responses, Chat Completions and Anthropic Messages endpoints, with fallbacks, Flex-tier routing, shadow testing and benchmark aliases.

- Full: https://www.anchorterminal.com/tools/ramp-router.md (~9,150 tokens) · this version ~1,930 tokens · JSON https://www.anchorterminal.com/tools/ramp-router.json · canonical https://www.anchorterminal.com/tools/ramp-router
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-10-28 · pricing https://www.anchorterminal.com/tools/ramp-router#pricing · reviews https://www.anchorterminal.com/tools/ramp-router#reviews
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

**C · 54.5/100 · rank #798 of 1218 · #21 in Model APIs & inference · not agent-ready · confidence medium**


Assessment: Detailed Markdown docs with typed error codes, per-model capability data and routing an agent can set per request. No OpenAPI document, status page, changelog or published rate limits were found. Signup needs a US business email in a browser, and Ramp records prompts and outputs for a year by default unless switched off.

## Facts

- Kind: Model router · vendor: Ramp · category: Model APIs & inference · legal entity: Ramp Business Corporation · provenance 83/100
- Endpoint: `https://api.router.com/v1` (HTTP)
- Auth: API key · pricing: Pay per use · x402: no · licence: Closed service under the Router Terms of Service (last updated 31 August 2026). The Ramp CLI that configures coding agents for Router is MIT.
- Probe metrics: not measured yet (probes haven't run)
- Surface graded: The gateway at https://api.router.com/v1, read from its docs, model pages, legal terms and site on 27 October 2026. No account was created and no call was made. An unauthenticated `GET /v1/models` answered 401 `invalid_api_key`. app.router.com is disallowed by robots.txt and was not read
- Endpoints: `GET /v1/models`, `POST /v1/responses`, `POST /v1/chat/completions`, `POST /v1/messages`, `POST /v1/messages/count_tokens`, `POST /v1/decisions` (limited availability) and `POST /v1/systemone` (TypeSafe Jev)
- Not available: Embeddings, files and batches. Chat Completions has no `models` fallback list, timeouts or `usage.cost`, and answers 501 for `logprobs`, `n` above 1, `seed`, `logit_bias` and legacy functions. Messages has no `models` list and no `usage.cost`
- Routing: Fallback lists of up to 15 `provider:model[:tier]` candidates on Responses. Cost-efficient Flex routing on by default, cross-provider routing by cost or throughput, Switchyard (NVIDIA NeMo) for coding agents, Jev Auto Routing (limited availability), benchmark aliases and shadow models, grouped in saved workloads
- Rate limits: No request or token limits published. The docs say rate limits vary by model. Technical limits are a 100 MiB request body, 15 candidates and 100 keys per user
- Errors: OpenAI or Anthropic error envelopes with a `code` per case, a `limit` object naming the spend cap and who set it, `x-request-id` and `x-trace-id` on every response, and `X-Router-Provider` naming the provider that answered. Provider `retry-after` headers are passed through
- Bring your own key: OpenAI, Anthropic, Google Gemini (required for Gemini), Fireworks, xAI, Baseten and Sail Research keys, and Amazon Bedrock through an AWS connection. Shared-key fallback is on by default and billed normally. Claude organisation billing is in limited availability
- Data handling: Ramp-side content recording on by default, kept for one year, switched off account-wide in Settings with asynchronous propagation and no deletion of existing archives. Flagged content may be kept up to 30 days even when storage is off. Shadow models need recording on
- Connectors: Beta. The Ramp CLI connects Claude Code, Codex and OpenCode to an MCP endpoint at https://api.router.com/v1/mcp for apps such as Slack, GitHub, Google Drive and Linear. Setup needs Sign in with Ramp and browser approval on every run, and some tools can write
- Company: Ramp Business Corporation, 28 West 23rd Street, New York, per the Router Terms of Service. A Router account is separate from a Ramp spend-management account
- Prices: Routing fee 0% percentage fee; Claude Opus 5.5 input $4 per 1M tokens; Claude Opus 5.5 output $20 per 1M tokens; GPT-6.1 Sol input $2 per 1M tokens; GPT-6.1 Sol output $10 per 1M tokens; Kimi K3 input $3 per 1M tokens; Kimi K3 output $15 per 1M tokens

- Scores: Reliability 25, Performance pending, Schema & documentation 58, Agent ergonomics 77, Security & auth 66, Payments & pricing 40, Task success pending, Maintenance & community 59, Transparency & trust 67 · total over the 7 assessed categories
- Why: Reliability, Hosted reading. · Schema & documentation, Router, scored on the API lines. · Agent ergonomics, Router, read on the model lines (tool use, structured output, caching, context, batch, SDKs, errors). · Security & auth, Router, read on the model lines. · Payments & pricing, No x402, MPP or L402 in any of the 41 docs pages. · Maintenance & community, Router, read on the model lines for a closed service. · Transparency & trust, Closed service under the Router Terms of Service of 31 August 2026, which bar reselling access and building a competing service.
- Sources: 23, open questions: 7, both in the full twin
- Capabilities: inference.router, inference.llm, inference.decision
- JSON: https://www.anchorterminal.com/api/v1/tools/ramp-router.json
- Verify (for the vendor): the badge `https://www.anchorterminal.com/badges/ramp-router.svg` or a link to https://www.anchorterminal.com/tools/ramp-router from a page on router.com or one of its subdomains, or the README of github.com/ramp-public/ramp-cli, then `POST https://www.anchorterminal.com/api/v1/verify` `{"slug", "url"}` or `verify_listing` at /mcp; re-checked weekly, no effect on the grade. Snippets in the full twin.
- Support: router@ramp.com; https://docs.router.com/resources/support-and-feedback; https://docs.router.com/resources/faq; more in the full twin

## Before you call it

1. Switch off content recording in Settings before sending private data. It is on by default and keeps prompts and outputs for a year
2. Read model ids from `GET /v1/models` at runtime. Ids are account-specific, an unknown one is a 404, and Gemini ids need your own Google key
3. Use `POST /v1/responses` with a `models` list for fallbacks. Router stops falling back once a stream has started
4. Set `fallback_timeout` and `provider_timeout`, or `allow_flex_tier: false`, when latency matters. Flex routing is on by default
5. Treat both `401 invalid_api_key` and `401 api_key_deactivated` as a possibly locked key, and stop retrying on `402`

## Connect

```bash
curl -fsSL https://router.com/install.sh | sh   # Ramp CLI, configures Claude Code, Codex, OpenCode, Pi, Hermes or Conductor
```

```bash
curl https://api.router.com/v1/responses \
  -H "Authorization: Bearer $RAMP_ROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"models":["openai:gpt-5.4-mini","fireworks:accounts/fireworks/models/kimi-k3"],"input":"Reply with one word: pong.","max_output_tokens":16}'
```

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| Vercel AI Gateway | B | 66.8 | inference.router, inference.llm, inference.decision | https://www.anchorterminal.com/tools/vercel-ai-gateway.min.md |
| RouterPlus | D | 51.3 | inference.router, inference.llm, inference.decision | https://www.anchorterminal.com/tools/routerplus.min.md |
| OpenRouter | B | 68.6 | inference.llm, inference.router | https://www.anchorterminal.com/tools/openrouter.min.md |
| Cloudflare AI Gateway | B | 66.9 | inference.llm, inference.router | https://www.anchorterminal.com/tools/cloudflare-ai-gateway.min.md |
| Granite Switch 4.1 | C | 59.4 | inference.decision, inference.llm | https://www.anchorterminal.com/tools/granite-switch.min.md |


## Panel reviews (0, desk reviews from public material, no calls made)