diff --git a/ra_aid/__main__.py b/ra_aid/__main__.py
index 7daa531..9d716b5 100644
--- a/ra_aid/__main__.py
+++ b/ra_aid/__main__.py
@@ -42,6 +42,8 @@ from ra_aid.config import (
     DEFAULT_TEST_CMD_TIMEOUT,
     VALID_PROVIDERS,
 )
+from ra_aid.database.repositories.key_fact_repository import KeyFactRepository
+from ra_aid.model_formatters import format_key_facts_dict
 from ra_aid.console.output import cpm
 from ra_aid.database import (
     DatabaseManager,
@@ -59,7 +61,7 @@ from ra_aid.prompts.chat_prompts import CHAT_PROMPT
 from ra_aid.prompts.web_research_prompts import WEB_RESEARCH_PROMPT_SECTION_CHAT
 from ra_aid.tool_configs import get_chat_tools, set_modification_tools
 from ra_aid.tools.human import ask_human
-from ra_aid.tools.memory import _global_memory
+from ra_aid.tools.memory import _global_memory, get_memory_value
 
 logger = get_logger(__name__)
 
@@ -523,6 +525,8 @@ def main():
                         ),
                         working_directory=working_directory,
                         current_date=current_date,
+                        key_facts=format_key_facts_dict(KeyFactRepository().get_facts_dict()),
+                        key_snippets=get_memory_value("key_snippets"),
                         project_info=formatted_project_info,
                     ),
                     config,
diff --git a/ra_aid/prompts/chat_prompts.py b/ra_aid/prompts/chat_prompts.py
index 042864c..a914b5a 100644
--- a/ra_aid/prompts/chat_prompts.py
+++ b/ra_aid/prompts/chat_prompts.py
@@ -11,6 +11,15 @@ from ra_aid.prompts.web_research_prompts import WEB_RESEARCH_PROMPT_SECTION_CHAT
 # Chat mode prompt for interactive mode
 CHAT_PROMPT = """Working Directory: {working_directory}
 Current Date: {current_date}
+
+<key facts>
+{key_facts}
+</key facts>
+
+<key snippets>
+{key_snippets}
+</key snippets>
+
 Project Info:
 {project_info}
 
@@ -88,6 +97,7 @@ You have often been criticized for:
     - Not calling ask_human at the end, which means the agent loop terminates and dumps the user to the CLI.
     - Not calling tools/functions properly, e.g. leaving off required arguments, calling a tool in a loop, calling tools inappropriately.
     - If the user asks you something like "what does this project do?" you have asked clarifying questions when you should have just launched a research task.
+    - Doing too many research tasks when it could all be done with a single request_research_and_implementation call.
 
 <initial request>
 {initial_request}