WRITER

What it is

WRITER is an enterprise agent platform sold to business teams, not to engineering. A marketer or revenue-ops lead describes a task in plain language, and WRITER turns it into an agent that runs on the company’s own content, data and brand rules, under controls IT sets once. It runs on WRITER’s own Palmyra models rather than renting someone else’s, and since September 2026 it markets itself explicitly as the agent platform for enterprise marketing and revenue teams.

It is not a general chat assistant like ChatGPT, and it is not an enterprise search engine like Glean. It is closest to a governed workshop where non-engineers build repeatable agents. WRITER is a San Francisco private company that raised a $200M Series C at a $1.9B valuation in November 2024; named customers include Accenture, Intuit, Qualcomm, Uber and Vanguard.

What changed in 2026

• Agent Skills and Playbooks (25 March 2026). A user describes a workflow, or pastes an existing process document, and WRITER converts it into an executable Playbook that pulls in Skills, connectors, knowledge and brand voice. It shipped with a library of more than 200 prebuilt enterprise Skills; WRITER says customers had already deployed over 5,000 Playbooks.
• Event-based triggers (30 April 2026). Agents now start on a signal instead of a prompt — a new Gong call, a Gmail message, a calendar event, a file landing in Google Drive or SharePoint, a Slack post.
• Palmyra X6 (13 August 2026). A new flagship model built on the open-weight GLM 5.2, priced at $2 per million input tokens and $8 per million output. WRITER says X6 plus harness changes cut the average cost of an agent task by 52% and made tasks 48% faster in its own tests.
• Enterprise Brain (9 September 2026). A shared context layer that maps accounts, opportunities, campaigns, content and customer conversations, with team-level Agent Memory, plus WRITER Meet for Zoom, Teams and Google Meet.

Connectors run through a WRITER-built MCP gateway that checks identity and permissions per action, and IT can add internal systems as custom connectors from an OpenAPI spec or an MCP server configuration. Named connectors include Salesforce, HubSpot, Gong, ZoomInfo, Slack, Snowflake, Databricks and Adobe Experience Manager.

A use case worth buying it for

Post-call follow-up for an enterprise sales team. A Gong recording lands, a trigger fires, and a Playbook drafts the recap email in the company’s approved voice, pulls the account’s open opportunity from Salesforce, updates next steps, and posts the draft to the rep in Slack for approval. The ops lead who owns the process builds and edits it without filing an engineering ticket, and the brand and compliance rules apply to every draft without the rep having to remember them.

Pricing

• Starter — $29 per user a month billed annually, or $39 monthly; up to 5 users, 5 Playbooks, 3 active connectors, 3 scheduled routines, and a limited Knowledge Graph. 14-day trial, no card.
• Enterprise — custom. Unlimited Playbooks, connectors and routines, SAML SSO, departmental voice profiles and advanced guardrails.
• API (AI Studio) — Palmyra X6 at $2/$8 per million input/output tokens; Palmyra X5, with a 1M-token context window, at $0.60/$6.

Starter is a trial tier; anything past five people is an Enterprise quote. Vendr’s transaction data puts the median WRITER contract at about $28,900 a year, 100–500-user deployments at $75K–$250K, and 500-plus-user deployments above $500K. Vendr also reports total cost running 20–40% above the subscription once API overages, services and support are added.

Best for

Marketing-ops and revenue-ops leaders at companies of 1,000 or more people who need governed, on-brand agents for content, account research and follow-up work, and who want business users — not engineers — to own the Playbooks.

Alternatives, and when to pick them instead

• Microsoft 365 Copilot and Agent 365 — the market-share default. Pick it when your work lives in Microsoft 365 and Copilot at $30 per user a month is already in the budget.
• Glean — the enterprise search leader. Pick it when the core problem is finding answers across dozens of systems, not producing branded output.
• Salesforce Agentforce — pick it when the agent’s job starts and ends inside Salesforce records and you’d rather pay per action than per seat.
• Dust — the fast-growing, transparent-price alternative at about $31 per user a month with no seat minimum. Pick it for a mid-market team that wants a no-code agent builder without an enterprise sales cycle.
• Claude Cowork — pick it when a frontier model’s reasoning matters more than brand governance and you can supply the guardrails yourself.

Watch-outs

• The model is WRITER’s, and the choice is narrower. Palmyra X6 scores well in WRITER’s own evaluations, but those are vendor benchmarks. Guard: run your ten hardest real tasks through Palmyra and through the frontier model you’d otherwise use during the trial, and write “full interoperability with your systems and LLMs” — the Enterprise tier’s promise — into the contract as a named, supported model list.
• Starter is not a real entry price. The five-user cap and three-connector limit mean a department rollout jumps straight to a custom quote. Guard: get the Enterprise per-seat price, minimum seat count and API overage rates in writing before the pilot, and use Vendr-style benchmarks to push back.
• The product is being repositioned fast. Four major launches in six months, and a narrowing of the pitch from every business function to marketing and revenue teams. Guard: buy for the Playbooks and connectors you can test today, and put Enterprise Brain and Agent Memory in the pilot, not in the business case.
• Triggers act without a human. An event-based agent that emails a customer on a bad signal does damage at machine speed. Guard: keep an approval step on anything that reaches a customer or writes to the CRM until the Playbook has run cleanly for a month.