
Argent with Anthropic
Route Claude traffic through Argent. Every call to Anthropic is checked against your approved models, scanned for sensitive data, held against a budget and signed, and the Anthropic key never reaches the agent.
What Argent does with Anthropic
Argent governs every call to the Anthropic Messages API, from your own apps and from coding agents such as Claude Code.
- Messages API: Argent reads Anthropic's own request format, so model rules, data protection and metering work on what Claude will actually see.
- Approved models only: a Claude model nobody approved is refused before the request leaves.
- Model Swap across providers: Argent translates between the Anthropic, OpenAI, Gemini and Bedrock formats, streaming and tool calls included, so a swap can cross providers.
- Hosted tools under control: provider-hosted tools such as web search, web fetch and code execution are refused unless your tenant allows that class.
- Prompt caching priced right: cache reads and writes, 5-minute and 1-hour, are priced at their own rates.
- Coding agents: Claude Code is tested end to end through Argent's TLS-inspecting proxy.
Anthropic retires models on a published schedule, and requests to a retired model fail (Anthropic model deprecations). With a model class, one change in Argent moves every agent to the replacement.
How it connects
One endpoint, one key in your vault, and every agent points at Argent instead of Anthropic.
- Add the endpoint once. On the AI Providers page, one action creates the Anthropic endpoint with its credential. The Anthropic API key is written to your vault and Argent keeps only the vault reference.
- Register each agent with an owner. Every agent or app gets its own Argent key and a named human owner. Its Connect tab shows copy-paste setup and a VERIFY CONNECTION check.
- Point it at Argent. The agent's base URL becomes its route on the Argent gateway,
/<tenant>/<endpoint>/, and the Argent key takes the place of the Anthropic API key. Claude Code connects through the TLS-inspecting proxy. - The gateway swaps keys on the way out. It removes the agent's Argent key, fetches the Anthropic API key from your vault and attaches it at egress, after checking Anthropic's certificate.
- Rotate in your vault. Every gateway replica drops the cached key, and no agent changes.
Your vault can be HashiCorp Vault or OpenBao, AWS Secrets Manager, CyberArk Conjur or Infisical. The agent never holds the Anthropic API key, so a leaked agent key is useless anywhere but Argent, and you can switch it off in one place. See Keyless Agents.
What you configure
Everything below is set once in the Argent console and applies to every agent that calls Anthropic through the gateway.
| Setting | Where | What it does |
|---|---|---|
| Approved models | Policy > Model and endpoint access | Default deny: a call is allowed only when the model is on the approved list. The list is platform AND tenant AND agent, and each layer can only narrow the one above. |
| Access rules | Same page | Allow or deny by provider, route type, model, hosted tools, function tools and output type, for the tenant, a department, a person or one agent. The narrowest rule wins, and a 7-day impact preview shows what a draft would change. |
| Model Swap | Model Swap picker | Runs an approved model in place of the one the agent asked for, including when a budget runs low, with no change to agent code. |
| Model classes | Policy > Model classes | A name such as fast stands for an ordered list of approved models, with failover on connection errors, 5xx, 429, timeouts and retired models. |
| Sensitive data | DLP | Per entity type: redact and record, record only, or block the request. |
| Budgets and caps | Budgets | Budgets for the company, org, department, person and agent. Each request is priced and held against the budget before it leaves, and an emergency cap stops a runaway agent now. |
| Mode | Per control | Off, watch or enforce. Watch records what would have been refused and changes nothing on the wire. |
Start in watch, read what would have been refused, then switch on enforce. See Watch Before You Enforce.
One request, step by step
What happens between your agent and Anthropic, in order.
- The agent sends its request to its Argent route, carrying its own Argent key and nothing from Anthropic.
- Who is calling. Argent identifies the agent from its key and checks it may use this endpoint. The endpoint comes from the route the gateway matched, never from a header the caller sets.
- Which model. Argent reads the model from the request body and checks it against the approved list and access rules before anything leaves.
- Swap or class. If a Model Swap rule or a model class applies, an approved model runs instead, and Argent translates the request and reply between provider formats where needed.
- Sensitive data. The request is scanned by a reader that knows the Messages format, so only text Claude will see is checked, and secrets and personal data are redacted, recorded or blocked under your policy. If the scan cannot run, the request is refused, never passed as clean.
- Budget. The request is priced and the amount is held against the agent's budget, so parallel calls cannot overshoot it.
- Key. The gateway attaches the provider key from your vault at egress.
- Reply and record. Token usage is read from the reply, the cost is settled, and the signed record is written.
What the record shows
Every Anthropic call leaves one signed record, allowed or refused.
- One row per call: the agent and the person who owns it, the tenant, the model, the decision and the policy that made it, latency, and the cost, token and cache figures.
- Tamper-evident: each row is hash-chained to the one before and signed with an ed25519 key held in Vault Transit, so an edited or missing row shows.
- Real cost: cost comes from the usage block in Anthropic's own reply, priced from your price book in USD per million tokens.
- No unpriced spend: an agent under a budget that calls a model with no price is refused, so a new model cannot slip past a cap.
Conversations can be opened and replayed from the record. See Session Replay and One-Click Audit Evidence.
Related capabilities and goals
Where to go next.
- Capabilities: Model Control, Sensitive Data Redaction, AI Spend Caps, Session Replay.
- Goals: Model control and routing, Budget and token spend, Agent identity and ownership.
- Works with: Claude Code, Anthropic SDK, Claude Agent SDK, Amazon Bedrock, OpenAI.
- Everything else Argent connects to: All integrations.
Download files
- PDFArgent Model Control data sheetApproved models, Model Swap, model classes and failover. One page.
- PDFArgent Keyless Agents data sheetProvider keys stay in your vault; agents carry only their own Argent key. One page.
- PDFArgent Sensitive Data Redaction data sheetPersonal data and secrets found and redacted before a prompt leaves. One page.
Want this running on your own use cases?
All downloads