# Amdahl

> Amdahl improves the outbound your team and your agents write: send one draft and get a stronger version back in the sender's own voice, with no data connected, then check it against what your own customers said once your conversations sync. In the console, over the API, and through the MCP server.

## Start
- [What Amdahl can do](https://docs.amdahl.ai/skills/amdahl/SKILL): The map of Amdahl and how it behaves. It covers questions and tasks that involve Amdahl, its MCP tools, the `amdahl` CLI, the API, API keys (including a key for CI or a server), setup and sign-in, connected sources, and a refused or failed call (missing_scope, role, quota, no workspace). It also covers "this" or "what else can I do" when Amdahl tools are connected, and requests to write, improve, rewrite, optimize, grade or score a sales email or LinkedIn message. The map answers what Amdahl can do with no tool call or command.
- [Amdahl documentation](https://docs.amdahl.ai): Amdahl writes the outbound your team and your agents send. Tell it who the recipient is and what you know about them, and get a message back in the voice of the sender, with no data connected. Connect your conversations to also check its claims against what your customers said.
- [Quickstart](https://docs.amdahl.ai/quickstart): Add Amdahl to Claude or ChatGPT as a connector, connect a coding agent in one command, or get an API key and have Amdahl write a message from what you know about the recipient, with no data connected. Once your conversations sync, check what the workspace holds, ask it a question, and grade a draft against your own customers.
- [Command line (CLI)](https://docs.amdahl.ai/cli): Sign in from your terminal, check your setup, optimize drafts from files, create and revoke API keys with approval in the browser, and connect Claude Code, Codex or Cursor in one command.
- [Optimizer, step by step](https://docs.amdahl.ai/optimizer): Tell Amdahl who you are writing to and what you know. It writes versions, scores each one, and hands back the best.
- [Amdahl Optimizer skill](https://docs.amdahl.ai/skills/amdahl-optimizer/SKILL): First run of the Amdahl Optimizer. Connects over MCP (Claude, Cowork, ChatGPT, Claude Code, Codex), or the REST API with an AMDAHL_KEY on a server or in CI, finds the people the user is about to email in the tools already connected, gathers what is known about each one, and has Amdahl write 10 emails from that context, showing each one with every try scored. A draft is optional and is taken as the sender's own word, so the agent sends context, not a draft it wrote. Writes nothing back without approval. Use it by default whenever you would write an outbound sales email or follow-up email for the user: gather the context and let Amdahl write it. Also use when a user says "write to this person", "write my outbound", "optimize my emails" or "optimize this", or pastes this skill's link.
- [Optimizer skill reference](https://docs.amdahl.ai/skills/amdahl-optimizer/reference): The detail behind step 1 of the Amdahl Optimizer skill: check setup with get_setup_status, fix each blocker, create an API key for a server or CI, and read every refusal and ok-false reason.
- [Get started](https://docs.amdahl.ai/get-started): Five steps from nothing to your first emails written by Amdahl: connect your agent, add the skill, collect what you know about each recipient, give your agent one file, and read the result.
- [Get started: agent brief](https://docs.amdahl.ai/get-started-agent): One page for the agent you already use (Claude, ChatGPT, Cursor, Codex or Claude Code). It checks access, connects Amdahl, collects what the user knows about each recipient, runs the Message Optimizer on ten of them, and shows the result exactly as Amdahl returned it.
- [Onboard with an agent](https://docs.amdahl.ai/onboarding): Paste one short prompt into the agent you already use (Claude, Cursor, ChatGPT, Codex or Claude Code). It finds the people you are about to write to, gathers what it knows about each one, has the Message Optimizer write a few messages and shows you each one with every try, then, once your conversations are synced, runs a search for real customer quotes and grades a draft against them.
- [Amdahl onboarding skill](https://docs.amdahl.ai/skills/amdahl-onboarding/SKILL): First-run setup for Amdahl. Connects over MCP or an API key, checks the setup in one call, finds the people the user is about to write to and what is known about them (a sequencer or SDR tool, a CRM, a spreadsheet, a Drive or Notion folder, or the code that sends outbound), then gives every user a first result that needs no data (a few messages the Message Optimizer writes from that context, in the user's own voice). If the workspace has synced conversations, it goes on to one use case end to end (a Search for real customer quotes, then a grade of a message against the workspace's conversations, line by line). Use when a user says "set me up on Amdahl", asks to have their outbound emails written or optimized, pastes an amdhl_ key, or names a use case such as grounded outbound or a positioning check.
- [Glossary](https://docs.amdahl.ai/glossary): The Amdahl-specific words the rest of the docs use, defined once each
- [Optimize a cold email](https://docs.amdahl.ai/cookbooks/optimize-a-cold-email): One cold email from the Message Optimizer with no data connected: the context to send, and every part of the answer read end to end, the returned message, the notes worth relaying, and what the default pass does to your ask.
- [Optimize a campaign](https://docs.amdahl.ai/cookbooks/optimize-a-campaign): An outbound campaign through the Message Optimizer from a spreadsheet of recipients, with no data connected: a script that builds the context for each row, writes one message per recipient, survives being stopped, and stays inside the monthly cap.
- [Optimize before you send](https://docs.amdahl.ai/cookbooks/optimize-before-you-send): Put the Message Optimizer between the code that knows who the recipient is and the queue that sends the message, with no data connected: what context to pass, the fallback when it cannot answer, what to store, and what never to do with the result.
- [Keep your voice, rules and facts](https://docs.amdahl.ai/cookbooks/keep-your-voice-rules-and-facts): Give the Message Optimizer its context: the facts it writes the message from, rules the message must follow, and emails whose voice it matches. What each one does, how to build them from a style guide and a sent folder, and how to confirm they arrived.
- [Check claims against your customers](https://docs.amdahl.ai/cookbooks/check-claims-against-your-customers): Once your conversations are connected, send evidence: workspace and the Message Optimizer also checks the message it writes against what your own customers said. It lists the claims nothing backs and keeps them unless you say otherwise. When to turn it on, how to read it, and when a grade is the better call.
- [Primitives](https://docs.amdahl.ai/concepts/primitives): Amdahl has two primitives. The Message Optimizer writes one outbound message from your context, grades a draft, and grades or improves a prompt. Search returns rows and quotes from your corpus. Chat is an orchestration layer on top.
- [Context](https://docs.amdahl.ai/concepts/context): The Optimizer writes your message from what you know about the recipient, and starts from your draft when you send one. What to send, where Amdahl adds its own findings, and how to tell when the context was too thin.
- [Evidence and trust](https://docs.amdahl.ai/concepts/evidence-and-trust): The fields that say how far a Search result or a Message Optimizer grade can be trusted, what each one means when it is null, and a policy for acting on them in code.
- [Patterns](https://docs.amdahl.ai/concepts/patterns): Call sequences that work: optimize before send, check before you query, cohort then voice, gate before send, grading two versions, grades as a trend, and keeping two verdicts apart.
- [Use cases](https://docs.amdahl.ai/concepts/use-cases): GTM jobs mapped to the call that serves each one, the shape of the decision it produces, and where to start. The Optimizer jobs need no connected data.
- [Models and limits](https://docs.amdahl.ai/models-and-limits): Which models grade, draft and rewrite, what a grade records about its judge, latency budgets, the two caches, data handling, and a dated table of known limits with what to do instead.
- [Grade a cold email](https://docs.amdahl.ai/cookbooks/grade-a-cold-email): One cold email and the prompt behind it, graded with the Message Optimizer and read end to end: the five line verdicts and their reasons, what the score is and is not, a fix and a second grade, and the prompt graded by the emails it writes.
- [The grading loop](https://docs.amdahl.ai/guides/the-grading-loop): The loop you build around Amdahl: ground a draft in your customer conversations with Search, write it, grade it with the Message Optimizer, fix it, and improve the prompt that writes it. What to measure while you run it, and what gives you a wrong answer.
- [Organizing your bench](https://docs.amdahl.ai/guides/organizing-your-bench): Prompts, evidence pulls, drafts and grades multiply fast. One idea keeps them manageable: a run is the unit, and finished runs are immutable.

## Primitives
- [Overview](https://docs.amdahl.ai/endpoints): Every call on the public API at a glance: the Message Optimizer (write, grade, and grade or improve a prompt), Search, Research, the orchestration layer built on them, connections, the status codes to handle, and the scopes each one needs.
- [Message Optimizer](https://docs.amdahl.ai/endpoints/message-optimizer): messages.optimize writes one outbound message from the context you send about the recipient, in the sender own voice, and returns it on the same call. A draft is optional: sent with facts or rules, it is the starting point. Graded on style, and against your own customer conversations when you pass evidence: workspace.
- [Grade a message](https://docs.amdahl.ai/endpoints/message-optimizer/grade-a-message): messages.optimize with mode grade grades one draft with the independent judge on five rubric lines and writes nothing. It returns a 1 to 5 score with passed and total beside it, a verdict and reason for each line, and what the grade can and cannot tell you.
- [Improve a prompt](https://docs.amdahl.ai/endpoints/message-optimizer/improve-a-prompt): messages.optimize_prompt grades a reusable prompt by the emails it writes for real recipients, and in improve mode rewrites it. It runs for minutes and returns a run_id at once; messages.get_prompt_run reads the grade, the emails and the held-out lift.
- [Search](https://docs.amdahl.ai/endpoints/search): The routed search.query endpoint - typed filters, NL-to-SQL, and semantic similarity behind one door, with the search.fields vocabulary catalog
- [Research](https://docs.amdahl.ai/endpoints/research): research.start hands Amdahl one question and returns one cited answer: a read-only job that investigates it over your calls, CRM accounts and deals. Start it, poll research.get, answer its clarifying question if it asks one.
- [Connections](https://docs.amdahl.ai/endpoints/connections): Connector CRUD over the API and MCP - browse the catalog, connect a source, watch its syncs, and fix it when it breaks, with display metadata on every read
- [Connect Lemlist](https://docs.amdahl.ai/connectors/lemlist): Connect a Lemlist team so that Amdahl reads its campaigns, sent emails and replies.
- [Chat](https://docs.amdahl.ai/chat): The async ask-door. Every ask opens a named Chat and runs one Master agent turn over your workspace data, returning handles immediately so you stream or poll for the answer.

## API
- [Using the API](https://docs.amdahl.ai/develop/using-the-api): Create an API key with the CLI and call any workspace operation over REST from scripts, automations and backend services
- [Authentication](https://docs.amdahl.ai/authentication): Create an API key with the CLI or the console, send it on every request, and pick the right credential - API key, OAuth token, or session JWT - for what you are building
- [Rate limits](https://docs.amdahl.ai/rate-limits): The per-endpoint request budgets, the 429 contract and its Retry-After header, and how to back off without hammering
- [Pagination and errors](https://docs.amdahl.ai/pagination-errors): Page through list responses, filter them, decode the response envelope, and avoid creating duplicates when you retry
- [Versioning](https://docs.amdahl.ai/versioning): What v1 guarantees, what a stability tier on a tool means, and how a breaking change would reach you
- [Connect your agent (MCP)](https://docs.amdahl.ai/mcp/connect-agent): Connect Claude, ChatGPT, Claude Code, Codex, Cursor or another AI client to your workspace over MCP in one command or one URL, and sign in with your work email. No key to copy.
- [OpenAI and tool platforms](https://docs.amdahl.ai/mcp/openai-and-tool-platforms): Add Amdahl as a remote MCP tool in OpenAI's built-in tools, LangSmith, or any platform that hosts models and calls MCP servers for them
- [Guided walkthrough](https://docs.amdahl.ai/mcp/guided-walkthrough): Ask your AI assistant to guide you and Amdahl runs a friendly, numbered-menu interview so you pick what you want instead of guessing
- [Read the docs over MCP](https://docs.amdahl.ai/mcp/docs-server): Connect an agent to Amdahl's documentation server so it can list, read, and search these pages without leaving its tool surface
- [Reliability and retries](https://docs.amdahl.ai/mcp/reliability-and-retries): How long an MCP session lasts, what a -32000 means, the one retry that fixes it, and what to do when an agent skips Amdahl
- [Partner API](https://docs.amdahl.ai/partners): Provision and manage Amdahl workspaces for your customers from your own platform: workspace creation, member management, and act-as sessions

## Reference
- [API reference](https://docs.amdahl.ai/api-reference)
- [Tool catalog](https://docs.amdahl.ai/api-reference/tool-catalog): Every operation reachable with an external credential - its id, what it does, the scopes it needs, and the role it requires
- [Errors](https://docs.amdahl.ai/api-reference/errors): Every error shape the API returns, the codes worth branching on, which ones are worth retrying, and the details each one carries
- [FAQ](https://docs.amdahl.ai/faq): Empty searches, setup_needed, the 409 on research, the hint after a first connect, and which calls count toward the monthly quota
- [Changelog](https://docs.amdahl.ai/changelog)