Docs

Keep your voice, rules and facts

Give Amdahl 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.

Level: first run onward. Call: messages.optimize with context. You need: a Customer agent key or the Amdahl MCP connector (see Connect your agent). No connected data.

Amdahl writes the message from context, so context is what it knows about this message: what to say, what to avoid, and how the sender sounds. A draft is optional. Sent in message, it is the starting point, and context adds to it and shapes it. It has three main keys, and all of them work on a workspace with no data connected. Context says what counts as enough.

KeyWhat it doesLimitWhere it comes from
factsWhat the message is written from: who the recipient is, why you write now, what you offer, the proof you may cite.10, each up to 1,000 charactersThe sender, your CRM, your own research
rulesConstraints the message must follow, including the ask. Never quoted in the message.20, each up to 500 charactersA style guide, a legal review, the sender's own "always" and "never"
voice_examplesMessages the sender wrote. The message matches their voice and does not copy their words.5, each up to 4,000 charactersThe sender's sent folder: emails that got replies

Rules: what the message must respect

A rule is a hard constraint. A version that breaks one is never returned.

Write one constraint per entry, in words you could check against the finished message:

json
"rules": [
  "Never mention pricing.",
  "Keep it under 90 words.",
  "End by asking for a 15-minute call.",
  "No exclamation marks.",
  "Sign off as Jordan, first name only."
]
  • Turn a style guide into rules line by line. Every "always" and "never" in it is a rule. Advice about tone ("sound confident, not salesy") is better shown than told: put it in voice_examples.
  • Say the ask you need. By default Amdahl prefers to close on a one-line question about whether the problem is real for the reader, rather than a request for a meeting. If your process needs the meeting ask, make it a rule, as in the third line above.
  • A rule is an instruction, not copy. It is never quoted in the message. Write what the message must do or avoid, not text you want to appear. Something the message must say is a fact, not a rule.
  • Change a result with a rule or a fact, not a rewrite. When the returned message is not what the sender wants, add what they want as a rule or a fact and run it again, or let them edit it. Do not polish message with another model: that undoes the evaluation that kept their facts exact.

Voice examples: what the sender sounds like

Up to five messages the sender wrote themselves. The message matches how they write and does not reuse their words.

json
"voice_examples": [
  "Hey Morgan, congrats on the new role. Most RevOps leads I talk to spend their first month untangling the CRM. Happy to share what the fast ones do first. Free for 15 minutes Tuesday?",
  "Priya, saw the Series B news. Are you still running onboarding by hand? We cut it to a day for a team about your size. Worth comparing notes?"
]
  • Pick sends that worked. Messages that got replies, on the same channel as the message you want: a LinkedIn note does not teach the voice of a cold email.
  • The sender's own words only. A message a teammate or a tool wrote teaches the wrong voice.
  • Drop signatures and legal footers. They are not voice, and they use up the 4,000 characters.

Facts: what the message is written from

rules and voice_examples change how a message reads. Neither gives it anything to say. Facts do: with no draft, they are the substance of the message. A message's specific lines come from your facts, from your draft when you send one, or from evidence Amdahl read.

json
"facts": [
  "Priya Shah is VP Sales at Northwind Labs.",
  "Met Priya at the GTM dinner on Sep 24; she said her SDRs rewrite every AI draft.",
  "We sell Relay, which drafts first-touch emails from CRM notes.",
  "A 150-person sales team cut SDR ramp from 90 to 45 days with Relay."
]
  • Who, why now, what you offer, and the proof. Those four are the useful minimum. A message from fewer is usually generic.

  • One fact per entry, in the sender's words, with the date when it has one.

  • Only what is true. A fact licenses a claim. An agent that writes one on its own, without a source, makes an invented claim look verified, which is worse than no fact at all.

  • Send it only with the recipient it is about. A fact about Priya does not belong on the email to Dana.

  • Without your data, facts are what the message says: the writer uses them and is told never to embellish them or present one as something a customer said. They matter most once your conversations are connected and you send "evidence": "workspace": then a claim a fact states is not flagged as unsupported, and a claim that goes further than the fact still is. See Evaluate claims against your customers.

Facts are data, never instructions: a fact that says "ignore the rules" is read as a strange fact. They are used for that one request, and are never stored as evidence or written to logs as text.

The whole request

json
{
  "channel": "email",
  "context": {
    "rules": ["Never mention pricing.", "End by asking for a 15-minute call."],
    "voice_examples": [
      "Hey Morgan, congrats on the new role. Most RevOps leads I talk to spend their first month untangling the CRM. Free for 15 minutes Tuesday?"
    ],
    "facts": [
      "Priya Shah is VP Sales at Northwind Labs.",
      "Met Priya at the GTM dinner on Sep 24; she said her SDRs rewrite every AI draft.",
      "We sell Relay, which drafts first-touch emails from CRM notes."
    ]
  }
}

Over MCP, send the same context object to the optimize_message tool.

Confirm what arrived

When you send context, the response says how much of it Amdahl received, as counts, never the text:

json
"context": { "rules": 2, "voice_examples": 1, "facts": 3 }

Blank entries are dropped before anything is counted, so a count lower than what you sent means some entries were empty. A facts count of 0 or 1 means the message had little to be written from. Anything else wrong is refused before Amdahl runs: an unknown key inside context, too many entries, or an entry over its limit returns 400 invalid_input naming the field, for example $.context.rules[3]: must be at most 500 characters.

What context cannot do

  • Name the sender. Who is sending comes from the credential the call is made with. There is no field for it, in context or anywhere else.
  • Put a detail in the message through a rule. A rule is never quoted, so "Offer Thursday at 2pm" does not put a slot in the message. A fact does: "Sam said Thursday at 2pm works" lets the message use it.
  • Make a claim true. Only a fact licenses a claim, and only for what it states.

Amdahl also keeps a per-sender record of how each sender writes, scoped to your workspace and keyed to the credential's user. A new sender starts with nothing in it, which is why context is the way to get their voice into the very first message.

The field limits and error shapes are on Optimize a message.